assessing the performance differences between hospitals
TRANSCRIPT
Seton Hall UniversityeRepository @ Seton HallSeton Hall University Dissertations and Theses(ETDs) Seton Hall University Dissertations and Theses
Winter 12-5-2016
Assessing the Performance Differences BetweenHospitals With and Without Meaningful Use ofElectronic Health Records on Care OutcomesJoseph G. [email protected]
Follow this and additional works at: https://scholarship.shu.edu/dissertations
Part of the Health Information Technology Commons
Recommended CitationConte, Joseph G., "Assessing the Performance Differences Between Hospitals With and Without Meaningful Use of Electronic HealthRecords on Care Outcomes" (2016). Seton Hall University Dissertations and Theses (ETDs). 2226.https://scholarship.shu.edu/dissertations/2226
ASSESSING THE PERFORMANCE DIFFERENCES BETWEEN HOSPITALS WITH
AND WITHOUT MEANINGFUL USE OF ELECTRONIC HEALTH RECORDS ON CARE
OUTCOMES
BY
Joseph G. Conte
Submitted in partial fulfillment of the
requirements for the degree of Doctor of Philosophy in Health Sciences
Seton Hall University
2016
2
Copyright © Joseph G. Conte 2016
3
Assessing the Performance Differences between Hospitals with and without Meaningful
Use of Electronic Health Records on Care Outcomes
By
Joseph G. Conte
Dissertation Committee:
Terrence F. Cahill, Ed.D., FACHE (Chair)
Fortunato Battaglia, MD, Ph.D.
Ning Zhang, MD, Ph.D.
Kenneth Ong, MD, MPH
Submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Health Sciences
Seton Hall University
2016
4
ACKNOWLEDGEMENTS
Many individuals deserve acknowledgement for their support and
contributions to the journey, the completion of this study and the achievement of
this Ph.D. First I would like to acknowledge my dissertation committee members.
Dr. Cahill, your untiring support throughout this process made the
difference in achieving this end result. The work is materially better for the many
hours spent challenging me to improve the end product. I am extremely grateful
for your guidance and fortitude. I know that the academic perspective of this
material is in great measure due to your guidance.
Dr. Battaglia, I am extremely grateful to have the benefit of your
experience in reviewing the clinical aspects of this paper in supporting the effort.
Your timely advice and willingness to contribute to its completion were very
important to its content.
Dr. Zhang, as the “big data” expert of this committee I am most
appreciative of your formative guidance in assembling the datasets and content.
Your expert guidance provided an invaluable focus on managing such a robust
and large data base.
Dr. Ong, as a colleague and friend your support has been invaluable in
pursuit of this Ph.D. Your stature in the health informatics community was a
5
constant reminder to me of how important this work was in putting a lens of
electronic health records.
A special thanks to Dr. Deluca and Dr. Zipp, both members of the original
faculty who interviewed me upon my application to the program. You are a
special example to all the students in your integrity, diligence and intellectual
rigor. I would also like to recognize Joann Deberto for her special help in
navigating the maze of University requirements at each step of this long process.
To all my friends who have supported and encouraged me through this
arduous and sometimes uncertain path, it is with great pride I can say your trust
was not misplaced.
Lastly, and most important I would like to thank my family members who
have been dragged along through this process. Through thick and thin you have
never been anything but a great cheer leading section and understanding of the
sacrifice that was needed to cross the finish line. When life intrudes on your path
it is those closest to you that keep everything in balance.
6
DEDICATION
This dissertation study is dedicated in memory of my sister Anita Rose
Conte. A formidable personality and one of the most impactful people in my life.
Her perseverance and dedication to those things that she held to be important is
a constant reminder that only the best is good enough. Gone far too soon but
never forgotten.
7
Table of Contents
Acknowledgements ......................................................................................................... 4
DEDICATION .................................................................................................................. 6
LIST OF TABLES ............................................................................................................ 9
LIST OF FIGURES ........................................................................................................ 11
ABSTRACT ................................................................................................................... 13
1. INTRODUCTION ..................................................................................................... 16
Background of the Problem ............................................................................... 16 Purpose of the Study .......................................................................................... 21
Study Variables .................................................................................................. 23
Research Questions ........................................................................................... 26
Significance of the Study .................................................................................... 27
Operational Definitions ....................................................................................... 28
Conceptual Model ............................................................................................... 28
2. LITERATURE REVIEW ........................................................................................... 34
Value Based Purchasing ................................................................................... 38 Process versus Outcome .................................................................................. 42
Summary of the Literature to Date ..................................................................... 46
Gaps in the Literature ........................................................................................ 47
3. RESEARCH METHODS .......................................................................................... 50
Research Design ............................................................................................... 50 Sample .............................................................................................................. 50
Description of Study Variables ............................................................................ 52
Data Collection ................................................................................................... 54
Data Analysis ...................................................................................................... 55
8
Data Analysis Methods………………………………………………………………..57
4. RESULTS ................................................................................................................ 59
Introduction ......................................................................................................... 59 Characteristics of the Sample ............................................................................. 59
Dependent Variable Results ............................................................................... 61
Results by Dependent Variable .......................................................................... 63
Cost per Discharge ............................................................................................. 78
Summary ............................................................................................................ 80
5. DISCUSSION........................................................................................................... 83
Background ........................................................................................................ 83
Discussion .......................................................................................................... 85
Theoretical Implications ...................................................................................... 92
Limitations .......................................................................................................... 94
Future Research ................................................................................................. 95
Closing Comments – Future Industry Trends ..................................................... 97
References
Appendix A- Institutional Review Board (IRB) Approvals
Appendix B- Definitions
9
LIST OF TABLES
Table I. Study Variables…………………………………………………………24
Table II. Dependent Variables………………………………………………….25
Table III. Examples of Indicators Related to Structure, Process &
Outcome………………….………………….………………….…………………45
Table IV. Frequencies and Percentage of Total by Categorical Variables…60
Table V. Mean Performance Data by Care, Cost and Safety Variables……62
Table VI. T-test Results on Quality Variables by Meaningful Use Status…..63
Table VII. T-test Results on Heart Attack Mortality by Meaningful Use
Status………………………….………………….………………………………...64
Table VIII. T-test Results on Heart Attack Readmission by
Meaningful Use Status……………………………………………………………66
Table IX. T-test Results on Heart Failure Mortality
by Meaningful Use Status………………………………………………………..68
Table X. T-test Results for Heart Failure Readmission by
Meaningful Use Status……………………………………………………………70
Table XI. T-test Results on Pneumonia Mortality
by Meaningful Use Status………………………………………………………..72
Table XII. T-test Results on Pneumonia Readmission
by Meaningful Use Status……………………………………………………….74
Table XIII. T-test Results on Composite Safety
10
Score by Meaningful Use Status………………………………………………76
Table XIV. T-test Results on Standardized Cost Per Discharge Metric by
Meaningful Use Status………………………………………………………….78
Table XV. Levene’s Test for Equality of Variance of Quality Variables
(2 sample t-test)……………………………………………………...................82
11
LIST OF FIGURES
Figure 1. Healthcare Spending Per Citizen Compared to US………………17
Figure 2. Healthcare Indicators - International vs U.S………………………17
Figure 3. Application of Donabedian Model Assessment of EHR………….30
Figure 4. Current State vs. Future State Evolution in VBP…………………32
Figure 5. The Four Medicare Regions………………………………………...56
Figure 6. Histogram of Heart Attack
Mortality Rate 2013-Hospitals (MU=NO) …………………………..65
Figure 7. Histogram of Heart Attack
Mortality Rate 2013-Hospitals (MU=YES) .………………………….65
Figure 8. Histogram of Heart Attack
Readmit Rate 2013 -Hospitals (MU=NO) …………………………..67
Figure 9. Histogram of Heart Attack
Readmit Rate 2013 -Hospitals (MU=YES) …………………………67
Figure 10. Histogram of Heart Failure
Mortality Rate 2013-Hospitals (MU=NO)…………………………..69
Figure 11. Histogram of Heart Failure
Mortality Rate 2013-Hospitals (MU=YES)………………………….69
Figure 12. Histogram of Heart Failure
Readmit Rate 2013-Hospitals (MU=NO)…………………………..71
Figure 13. Histogram of Heart Failure
12
Readmit Rate 2013-Hospitals (MU=YES)………………………….71
Figure 14. Histogram of Pneumonia
Mortality Rate 2013-Hospitals (MU=NO)…………………………..73
Figure 15. Histogram of Pneumonia
Mortality Rate 2013-Hospitals (MU=YES)…………………………73
Figure 16. Histogram of Pneumonia Readmit
Rate 2013 -Hospitals (MU=NO)……………………………………..75
Figure 17. Histogram of Pneumonia Readmit
Rate 2013 -Hospitals (MU=YES)……………………………………75
Figure 18. Histogram of Safety-Hospitals (MU=NO)………………………….77
Figure 19. Histogram of Safety-Hospitals (MU=YES)………………………...77
Figure 20. Histogram of Cost per Discharge-Hospitals (MU=NO)…………..79
Figure 21. Histogram of Cost per Discharge-Hospitals (MU=YES)…………80
13
ABSTRACT
Assessing the Performance Differences between Hospitals with and
without Meaningful Use of Electronic Health Records on Care
Outcomes
Joseph G. Conte
Seton Hall University, 2016
Dissertation Chair: Terrence F. Cahill, Ed.D. FACHE
Background and Purpose of the Study: The U.S. healthcare system at $3
trillion, is the sixth largest economy in the world. The federal government is the
largest purchaser of healthcare in the country. In the past decade it has been on
a quest to refocus its purchasing from volume to value. While spending nearly
double per capita than every other industrialized nation, U.S. healthcare
outcomes are consistently in the lowest quartile for every major indicator from life
expectancy to ambulatory sensitive conditions. The Crossing the Quality Chasm
Report (IOM) focused a lens on the dearth of electronic health record (EHR)
systems nationally. Resultant legislation, the HITECH Act, funded a $50 billion
investment to close this gap along with promulgation of standards known as
Meaningful Use (MU) to achieve interoperability. This investment and related
MU protocols for implementation warrant a careful examination to establish if the
intended improved outcomes have been achieved.
14
Methods: The study is a cross-sectional, retrospective design; it employs
two cohorts, Meaningful Use (MU) vs Non-MU hospitals. Publicly reported data
on clinical outcomes, cost and safety from 4221 or 95% of the nation’s hospitals
were included in the analysis to identify if there is a difference in outcomes
between the hospital cohorts.
Results: 2315 of the 4221 or 55% hospitals who were included in the
study met MU standards by 2013. The profile of an MU hospital was a non-
teaching (70%), geographically southern (40%), not-for-profit hospital (61%).
Non-Mu hospital had a similar profile, 78% non-teaching, 35% Southern and 60%
not-for-profit. Those hospitals who met MU had statistically lower mortality
(p<.05) rates for all three clinical conditions (heart attack, heart failure,
pneumonia) and statistically lower cost per discharge of $327 (p<.05). The
improved outcomes suggest a reduced cost of over $6 billion and 21,000 fewer
deaths.
Conclusion: The HITECH Act that committed over $50 billion in subsidy
incentive funds has dramatically increased EHR adoption nationally from 8% in
2009 to over 50% by 2013. The results from this suggest hospitals that had
implemented EHRs’ that meet MU standards demonstrate mortality and cost
outcomes that result in statistically significant cost and clinical care benefit.
16
CHAPTER 1
INTRODUCTION
BACKGROUND OF THE PROBLEM
The United States spends nearly double per capita what every industrialized
nation does on health care per the Organization for Economic Cooperation and
Development (Figure 1). However, ranks in the lowest quartiles of performance
for infant mortality, life expectancy, male and female healthy life expectancy and
nearly every major health indicator (Figure 2). These outcomes, coupled with an
abysmal patient safety record (IOM, 1999; Bates, 2001; Shekelle, 2011) underlie
systemic flaws in the system. The need to identify effective levers for change
began to evolve from government, industry and health economist in academia
(Porter, 2006). The gaps identified were not the absence of evidence based best
practices, clinical guidelines, competent practitioners or academic rigor but a
misaligned payment model and a dearth of electronic health records in the
nation’s hospitals (IOM, 2001).
17
Figure 2. Healthcare Indicators - International vs U.S.
In the search for solutions to the value inequity in American healthcare, the
report “Crossing the Quality Chasm: A New Health System for the 21st Century”
Figure 1. Healthcare Spending Per Citizen Compared to U.S.
18
(IOM, 2001), focused national attention on electronic health records (EHR). The
report highlighted structural shortcomings of the existing healthcare delivery
system, a major one being the “absence of real progress…toward applying
advances in information technology” (p. 115). The report stated that all healthcare
organizations should set goals for improvement, specifically that healthcare
should be: safe, patient-centered, efficient, effective, timely and equitable. In
support of the IOM findings, the Health Information Technology for Economic and
Clinical Health Act (HITECH) legislation was created to stimulate the adoption of
health information technology. This $50 billion investment, coupled with
excitement generated by literature (Hillestad, 2005) that suggested that nearly
$80 billion in savings would accrue from EHR adoption nationwide spurred a rush
to EHR adoption (Desroches, 2013).
While electronic health record systems (EHR) existed in hospitals for
decades, they functioned in isolation as departmental reporting and record
keeping tools. Laboratory, pathology, imaging, pharmacy and other ancillary
programs without interconnectivity left valuable information isolated and
inaccessible to multidisciplinary team users, resulting in excess utilization,
ineffective prescribing and safety lapses (Bates, 2001). An interconnected,
cohesive electronic health record platform with access to the most current
evidence base, clinical decision support and safety features to prevent errors was
lacking in over 92% of hospitals as late as 2008 (DesRoches, 2010).
19
In addition to the focus on the EHR gap, payment reform is another lever
being utilized to change the fundamental economics driving health cost, quality
and the wide variations in care utilization trends (Fisher, 2009). The endgame in
healthcare purchasing is not strictly one of cost control nor one of quality
improvement in isolation; it is a search for value (Porter, 2006; McHugh, 2010).
Value in healthcare is defined as health outcomes achieved per dollars spent
(Porter, 2006). Different models of payment reform such as Pay for
Performance, Value Based Purchasing, Accountable Care Organizations (shared
savings or risk models) and patient Centered Medical Home programs are all in
play in the search for sustainable models balancing patient centered care with
cost and quality outcomes (Eldridge, 2011). Ultimately, value-based purchasing
is part of a much broader policy “experiment” to advance value as a remedy for
spiraling health benefit costs and quality concerns in US healthcare. The
implementation of payment reform in parallel with the HITECH Act support for
EHR has created a naturally occurring experiment in which to study the
difference between hospitals that have adopted EHR technology and attested to
Meaningful Use (MU) standards versus those that have not. Meaningful Use is
the set of standards for EMR adoption and functionality as defined by the Office
of the National Coordinator (ONC). ONC not only defines the standards but also
governs the assessment process employed to evaluate eligible providers’ and
hospitals’ implementation progress to earn incentive payments.
20
The HITECH Act legislation committed a combined $50 billion in federal
and state funds to support the rapid expansion of EHR capability in the nation’s
hospitals and physician practices. The problem is that this legislation, while bold
and strategic, was designed without empirical evidence that the investment
would yield improved outcomes in population health, cost and safety. The belief
that EHR adoption represents a significant component of the solution to modulate
annual cost trajectory of healthcare and solve the quality and safety dilemma
rests on conflicting research which is at present inconclusive and depending on
measurement parameters, time periods included and available survey data
(Hillestad, 2005; Buntin, 2011; Blumenthal, 2010; Desroches, 2010; Rudin,
2014).
Further not only does the HITECH Act commit $50 billion in public funding,
the criteria to earn the subsidy payments require substantial investments by
hospitals and healthcare systems. To obtain subsidy funding, which began in
2011, healthcare providers had to demonstrate MU of electronic health records
(EHR). As stated above, MU is the set of standards for EHR adoption and
functionality as defined by the Office of the National Coordinator (ONC). ONC
not only defines the standards but also governs the assessment process
employed to evaluate eligible providers’ and hospitals’ implementation progress
to earn incentive payments. The opportunity to qualify for payments began in
2011 for hospitals attesting to meaningful use standards. In 2015, CMS began
withholding between 1 and 3% of Medicare payments to those hospitals that
21
failed to meet the MU requirements. It is estimated that the private funding
required for the nation’s hospitals could top well over $100 billion (Rand, 2005)
PURPOSE OF THE STUDY
This study was undertaken to focus a lens on the impact of EHR adoption,
a major health policy initiative under the HITECH Act (2009). HITECH was
implemented in parallel with Value Based Purchasing, the CMS program
designed to realign payment with value and away from the fee for service
structure focused on volume. This study is needed because these 2 policy
decisions created an intersecting impact on the healthcare system and have
created a major gap in the literature that this study was designed to address.
The purpose of this research therefore is to study whether there is a difference in
publicly reported outcomes of quality, safety and cost per discharge between
hospitals that have adopted EHR’s and achieved MU status versus those that
have not achieved MU in the era of value based purchasing.
As the healthcare payment system transitions to VBP, outcomes
measures focusing on quality and cost are the key indicators to assess
improvement and progress (VanLare, 2012; McHugh, 2010; Ranawat, 2009;
James, 2012). Prior studies focused solely on process measures for quality
measurement. These measures assess compliance with steps in care protocols,
while a proxy assessment of quality (Donabedian, 1988, 2005; Mant, 2001) they
do not align with current Value Based Payment models where hard outcomes
22
linked to expenditure such as readmission, cost per discharge, infection and
mortality rates drive reimbursement schema. Cost calculations were not
standardized and therefore not comparable from study to study (Himmelstein,
2009; Chaudry, 2006; DesRoches, 2010; Appari, 2009; Agha, 2011). This study
included safety, quality outcomes and cost per discharge measurements to
assess whether MU adoption of EHR systems makes a difference in these
outcomes.
Numerous studies (Himmelstein, 2009; Chaudry, 2006; DesRoches, 2010;
Appari, 2009; Agha, 2011) have demonstrated that EHR adoption has a generally
positive impact on process measures of quality. Most however, have found
negative to neutral outcomes on cost of care. Much of this research was
conducted prior to the creation of MU criteria (Himmelstein, 2009; Chaudry, 2006)
or standardized cost per episode of care. This gap indicates that this research
approach, outcome vs process is necessary to more closely focus a lens on how
MU standards can improve cost and value as organizational maturity and
experience with EHR evolves.
Therefore, in the context of value based purchasing (VBP), the secondary
purpose of the study is to utilize outcome measures such as mortality,
readmission and cost per episode of care linked to Value Based Payment models
as opposed to process measures to evaluate the differences in performance in
clinical quality and cost. These and other outcome measures are the foundation
of payment under the VBP methodology. The challenge to the validity of the
23
program is utilizing reliable quality indicators and a sound cost rubric (Wachter,
2006; Rudin, 2014, Kazley, 2009). Therefore, an underlying goal of this study is
to bring a level of consistent outcome measurement to the literature in the MU
era. This will stand in contrast to the inconsistent and complicated systems of
performance measurement previously applied when comparing results between
organizations outcomes and their MU status.
Study Variables
The independent variables selected for this research have been utilized
consistently in the literature focusing on EHR adoption (Chaudry, 2006; Jha,
2009; Himmelstein, 2009; Agha, 2011; Ding, 2011; DesRoches, 2013). The
variables are teaching status, case mix index (CMI) or acuity, region, discharge
volume and owner status (Table 1). Teaching status describes whether a
hospital participates in graduate medical education of physicians as defined by
Council of Teaching Hospitals and Health Systems (COTH). Teaching hospitals
characteristics include tendency toward more complex care, service for the
underprivileged, greater cost per discharge and urban locations (Shahian, 2014).
24
Table I: Study Variables
© 2016 Joseph G. Conte
Case mix is an extremely important variable in that it reflects the
complexity of care across various diagnostic related groups (DGR), but also
includes factors that account for regional variation in cost of services. DRG are
assigned by the primary reason for hospital care but also include a patient’s age,
sex race as well as co-morbidities. Co-morbidities are pre-existing medical
conditions that affect how care is provided and ultimately weigh on the total cost
of service and likelihood of a favorable outcome of care. CMI is represented as a
numeric value with a normalized base of 1.0. This represents the “average
hospital” CMI, a value greater than 1.0 reflects higher complexity of care and cost
of service, with the opposite holding true. In the VBP context CMI is a very
important hospital factor for reimbursement.
25
Region is a geographic construct coded 1-4 applied to mirror the 4 regions
recognized by the Center for Medicare coding system. Discharge volume
reflects total number of cases discharge from the hospital included deaths.
Obstetrical and pediatric cases are not included in Medicare calculations.
Ownership status falls into the three categories, not-for-profit, for-profit,
governmental.
In order to address cost, quality and safety measurements, the dependent
variables for this study are cost per inpatient discharge, morbidity and mortality
rates for heart attack, heart failure and pneumonia and a standardized patient
safety score (Table 2).
Table II: Dependent Variables
© 2016 Joseph G. Conte
26
Research Questions
These are the five research questions:
Is there a difference in clinical outcomes between hospitals that have
achieved Meaningful Use (MU) for their EHR systems verse hospitals that have
not achieved MU as measured by mortality and readmission rate for Heart
failure?
Is there a difference in clinical outcomes between hospitals that have
achieved Meaningful Use (MU) for their EHR systems verse hospitals that have
not achieved MU as measured by mortality and readmission rate for Heart
attack?
Is there a difference in clinical outcomes between hospitals that have
achieved Meaningful Use (MU) for their EHR systems verse hospitals that have
not achieved MU as measured by mortality and readmission rate for Pneumonia?
Is there a difference in clinical outcomes between hospitals that have
achieved Meaningful Use (MU) for their EHR systems verse hospitals that have
not achieved MU as measured by Agency for Healthcare Research and Quality
Patient Safety Indicator
Measurement?
Is there a difference in cost per discharge between hospitals that have
achieved Meaningful Use (MU) for their EHR systems verse hospitals that have
not achieved as measured by the CMS standardized cost per discharge metric?
27
Significance of the Study
There are over 4000 hospitals in the US confronted by what is essentially
a federal mandate to achieve MU adoption of their EHR systems, i.e. incur
penalties for non-adoption of 2% of gross Medicare revenue by 2016. Given the
magnitude of the investment required by hospitals and the $50 billion in
governmental subsidy available to transform healthcare in the US to an
interoperable EHR platform objective measurement of its efficacy is essential.
This study applies outcome not process measures that align with core VBP
indicators to evaluate whether EHR adoption under the MU guidelines can
achieve the promise of quality, cost and safety improvement highlighted in the
Crossing the Quality Chasm report (IOM, 2001).
In 2015, over $1.5 billion was withheld from hospitals that failed to achieve
the specified performance outcomes and then reallocated to hospitals based on
their overall performance on a set of outcome measures in clinical, cost and
satisfaction measures (www.cms.gov/VBP). The paradigm shift of moving away
from paying for volume of services to linking payment with outcomes for the 100
million Americans covered by Medicare is transformational. This is the most
significant change in US healthcare since government sponsored health
insurance came into being. Because commercial Medicare managed care plans
and Medicaid programs in each of the states are adopting this very same payment
methodology, the influence of value based purchasing in US healthcare is an
attempt at transformation on a grand scale (VanLare, 2012; Eldridge, 2011).
28
Operational Definitions
There are three main constructs used in this study that are identified in the
literature. These three constructs are Meaningful Use, Value Based Purchasing
and Quality/Outcomes Measurement. Meaningful Use (MU) refers to the criteria
developed by the Office of the National Coordinator (ONC) that defines the
interconnectivity standards and technical performance requirements necessary to
achieve MU designation under one of the three progressive levels to be eligible
for federal subsidy payments. Value-based purchasing (VBP), links payment
directly to the quality of care provided, it moves the focus from volume to value
(Porter, 2009). CMS has made VBP (www.cms.gov/Medicare/Quality-Initiatives)
the focus of its effort under the Affordable Care Act, to transform the current
payment system by rewarding providers for delivering high quality, efficient clinical
care based on value, not volume. The methodology most favored when
assessing quality of care in the hospital setting has been to use process
measures (Donabedian, 1988). The shift to outcome measures like mortality,
readmission, infection rates, etc. has occurred in parallel (VanLare, 2012;
Ranawat, 2009; Porter, 2009) with the need to shift from adherence to standards,
to risk adjusted outcomes of care where value not volume is the driver.
CONCEPTUAL MODEL
The measurement of outcomes is at the core of the analysis to determine
if there is a relationship between MU adoption of EHR systems and improvements
regarding quality of care and/or safety, and/or cost of care.” This study draws on
29
two theoretical frameworks, one focused on clinical care measurements and the
other focused on healthcare economics. The Donabedian Model is the dominant
paradigm for assessing the health care. It focuses on structure, process and
outcome to measure quality. The second is the Value Based Purchasing theory
advanced by an economist Michael Porter in his seminal work, Redefining
Healthcare (2006). The Porter model argues that the US healthcare system’s
inherently misaligned payment methodology resulted in the pursuit of high
volume, high margin services without a focus on the outcomes of health for the
population being served or the total cost of care incurred.
The successful adoption of EHR’s under the Meaningful Use (MU) criteria
developed by the CMS Office of the National Coordinator is based upon
achieving, both process and outcome milestones. The Donabedian model
therefore is uniquely suited to function as an overall theoretical model as it
incorporates the assessment of multiple process steps to achieve MU while at the
same time including the care outcomes as a core domain. How the domains of
EHR implementation processes and measures for the outcomes of quality and
cost of care align to form the theoretical measurement approach to assess the
difference in outcomes for MU and non-MU hospitals is illustrated in Figure 3.
30
Figure 3. Application of Donabedian Model Assessment of EHR
© 2016 Joseph G. Conte
Figure 3 illustrates the Donabedian structure, process, outcome matrix as
applied in this study to EHR adoption. Under structure, it encompasses the
physical facility, equipment, and human resources, as well as organizational
characteristics such as staff training and payment methods. The Process
elements include technical processes for EHR implementation, compliance with
ONC criteria under MU and provision of services influenced by EHR protocols.
And of course the outcome domain contains all the effects of healthcare on
patients or populations, including changes to health status, behavior, health-
related quality of life. This theoretical model, while oriented to process measures
31
as proxies for outcomes such as achieving MU status process criteria, is also as
flexible enough to incorporate outcome indicators into measurement criteria.
In fact, nearly thirty years ago Donabedian (1988) made an astute
observation that is at the core of the dilemma of the inexorable cost trajectory in
healthcare. He noted that “it is believed that as one adds to care, the
corresponding improvements in health become progressively smaller while costs
continue to rise unabated” (p. 1745). Consistent with current payment reform
focus on value, he postulated that it is possible to separate quality from
inefficiency by analyzing each added bit of expected usefulness against its
corresponding cost. Those providing care without regard to cost, he terms
maximalist. Those who provide care with a focus on weighing each additional bit
of expected usefulness against its corresponding cost, he terms optimalist.
Donabedian captures the essence of current healthcare reform debate by
focusing on the maximalist vs optimalist approaches to care and their respective
impacts on cost and health benefits. This value dilemma, excessive cost coupled
with inferior population health and safety outcomes, positions the US healthcare
system to search for systemic, effective solutions. What the Donabedian model
lacked was an economic framework to embody this concept into a system with
functional regulatory and financial underpinnings. This is where the work of Porter
with its emphasis on value, cost and outcome created the underpinning for the
value based purchasing paradigm.
32
Porter’s seminal work Redefining Healthcare (2006) focused on the
misaligned reimbursement model as the root cause for out of control cost and
poor health outcomes. He highlighted how the entrenched pay for volume rather
than payment for value system encouraged high cost, high volume care rather
than a focus on population health. In Figure 4. the change associated with
current to future state evolution in VBP is represented.
Figure 4. Current State vs. Future State Evolution in VBP
Source: New York State Value Based Roadmap, 2015
Porter also rejected reliance on compliance with process measures as an
effective measure system to govern reimbursement methodology preferring the
IHI inspired Triple Aim (IHI, 2006) outcome measures rubric. The current VBP
system informed by Porter’s work has shifted reimbursement from former pay for
volume to a focus on outcomes of care and total cost of service. After a transition
33
period, 2011-2013, where process measures dominated payment, outcomes of
care and cost now account for 90% of reimbursement for the program in 2016.
These two conceptual models, the Donabedian triad model of structure,
process and outcome, coupled with the Porter value based payment model are
the theoretical pillars adopted by this study to measure the impact of EHR
systems on hospitals. In order to bring a more detailed focus on the research to
date as well as the theoretical concepts and developments of EHR
implementation in the MU era, the literature on the topic is reviewed in the next
Chapter.
34
Chapter II
LITERATURE REVIEW
The early excitement on the potential quality and cost improvements benefits,
associated with EMR adoption was fueled by literature funded by the Rand Corporation
(Hillestad, 2005) that suggested that nearly $80 billion in savings would accrue from
EHR adoption nationwide. This estimate was based on savings associated with
efficiency, clinical decision support and safety improvements. When added to the
anticipated savings from improvements in preventative and chronic disease
management the estimated benefit increased to the $100 billion level.
Prior to the adoption of the HITECH Act (2009), which included over $50 billion in
federal Meaningful Use (MU) funding, the US seriously lagged other industrialized
nations in EMR adoption. In 2008, prior to the advent of MU incentives, fewer than 9%
of US hospitals had even basic EMR systems (DesRoches, 2013). The benefit of MU
incentives and threat of penalties (HITEC Act, 2009) jump started EMR adoption and by
2010 the proportion of US hospitals with basic EMR jumped to 15% (DesRoches, 2013).
Once financial incentives under MU began flowing in 2011, the adoption of EHR
systems nearly doubled to 27% (AHA IT Survey, 2013). The American Hospital
Association data (IT Survey, 2013) identified that EHR adoption was most robust in
large urban and teaching hospitals.
As the $30 billion HITECH Act (2009) subsidy to stimulate EHR into US
healthcare over the past 10 years has accelerated, the effort to assess its effectiveness
35
has grown. However, the literature (Chaudry, 2006; Himmelstein, 2009; DesRoches,
2010; Ding, 2011; Appari, 2012) that has investigated the association between the
investment in EHR and its effect on quality, safety and cost domains reveal mixed
results at best. Individual hospital, ambulatory practice and even health system
reviews illustrate improvements in domains such as medication safety (Bates, 2001;
Poon, 2012) or quality scores (Lindenauer, 2007), yet the results are far from
conclusive. Other recent studies have demonstrated little if any benefit in quality and
none in cost control (Ding, 2011; Himmelstein, 2009).
Appari, et al. (2012) reviewed HIT and quality data for a four year period (2006-
2010) for 3,921 non-federal hospitals. They measured quality by analysing process
measure compliance for pneumonia, heart attack and heart failure. Statistical analysis
was conducted using fixed effects linear panel regression models over a 5 year period,
2006-2010. Their study entitled "Meaningful Use of Electronic Health Record Systems
and Process Quality of Care: Evidence from a Panel Data Analysis of U.S. Acute-Care
Hospitals”, found adoption of EHRs did improve hospital process quality measures for
AMI, heart failure and pneumonia. This improvement was especially true for hospitals
that started with scores in the lowest quartile of performance. In an unexpected finding,
hospitals with EHRs that upgraded their basic systems to more advanced functionality
experienced a quality score decline. This finding prompted a cautionary conclusion
“technology implementation alone is not sufficient to produce quality improvement” (p.
17). As stated earlier, a limitation in this and other studies is that it was conducted
before MU criteria were codified by the Office of the National Coordinator. Their
36
evaluation of MU compliance was based on self-reported capabilities from the AHA
annual IT survey not assessment of MU compliance as required under the HITECH Act
(2009). Neither cost nor safety was included in the analysis critical factors when the
value of services are evolving into central theme in healthcare policy.
Spencer (2010) utilized a national cohort study based on primary survey data
about hospital EHR capability and publicly reported quality data. A regression analysis
was used to assess the relationship between EHR adoption and quality improvement for
heart attack, heart failure and pneumonia care. To evaluate the association between
quality improvement over time and the availability of an EHR, they compared hospitals
that maintained a system with those that reported having no system. The results were
striking for significant increase in quality scores for heart failure, less for heart attack
and none for pneumonia scores. Unlike the findings from the Appari, et al. (2012),
implementation of advanced systems did not result in decreased quality scores but did
result in smaller gains for AMI and heart failure.
Ding et al., (2011) examined the effects of EMR on the clinical, financial and
operational outcomes of U.S. hospitals. They utilized publicly reported data on EHR
adoption from 2006-2008 (the pre-MU era). The information was obtained from the
Health Information Management (HIMSS) database, the Hospital Quality Alliance for
quality scores and the American Hospital Association database for performance
metrics. The focus of the study was to test the effects of EMR adoption over time within
a hospital. This is a unique analysis moving away from EMR adoption as a binary
variable, i.e. EHR vs no EHR, to one looking at the effect of the adoption overtime.
37
Using simultaneous regression models they found that EMR adoption has a positive
and significant effect on cost and quality. Specifically, adoption history, the time an
organization was operational with an EMR, was associated with reduced cost per
patient day but not on length of stay. Therefore, overall cost per discharge were
essentially unchanged. The improvement in process quality measures for AMI, heart
failure and pneumonia clinical outcomes likewise increased over time. When comparing
the effect size, they found the impact on operational and financial outcomes more
significant than that on clinical quality measures.
Himmelstein (2009) linked data from an annual survey of computerization at
approximately 4,000 hospitals for the period from 2003 to 2007 with administrative cost
data from Medicare Cost Reports and cost and quality data from the 2008 Dartmouth
Health Atlas. Higher overall computerization scores correlated weakly with better
quality scores for acute myocardial infarction (r_0.07, P_.003), but not for heart failure,
pneumonia, or the 3 conditions combined. Utilizing multivariate analyses, more
computerized hospitals had slightly better quality. However, in comparing a hospital’s
overall computerization score, more computerized hospitals had higher total costs in the
2003-2007 period and a more rapid increase in computerization was associated with a
faster increase in computerization costs. Himmelstein, et.al (2009) concluded that as
currently implemented, hospital computing might modestly improve process measures
of quality but does not reduce administrative or overall costs. Further, hospitals that
increased their computerization more rapidly had larger increases in administrative
costs. As significant federal and state financial resources continue to be committed to
38
EHR adoption, the investments, as per Himmelstein, “rest on scant data” and “Recent
Congressional Budget Office reviews have been equally skeptical…” (pg. 2).
Value Based Purchasing
Payment reform is a lever being utilized to change the fundamental economics
driving health cost, quality and wide variations in care utilization trends (U.S Department
Health and Human Services, 2007; Eldridge, 2011). The endgame in healthcare
purchasing is not strictly one of cost control nor one of quality improvement in isolation;
it is a search for value. This means a movement away from payment for volume to one
of payment to reward quality outcomes that contain cost. Different models of payment
reform such as Pay for Performance, Value Based Purchasing, Accountable Care
Organizations (shared savings or risk models) and patient Centered Medical Home
programs are all active strategies in a search for sustainable models balancing patient
centered care with cost and quality outcomes (James, 2012; VanLare, 2012)
Ultimately, value-based purchasing is part of a much broader policy “experiment” to
advance value as a remedy for spiraling health costs and quality concerns in US
healthcare.
Regardless of the vehicle(s) chosen, until incentives to providers are aligned in
local or regional arrangements with population health as an ultimate measure of value,
the current siloed approach under fee-for-service will continue to promote perverse
resource utilization (VanLare, 2012). As demonstrated over the past 20 years by
Dartmouth Atlas reports (Fisher, 1999) regional variation in health care costs have no
correlation to differences in quality outcomes, acuity of care or cost of care delivery.
39
The report identified that in high cost regions (Fisher, 2009) utilization patterns by
physicians and others respond to the availability of high cost alternative services that
have no greater efficacy, including discretionary referrals to specialists. The payment
for volume of services provided which was the predominant model by which
discretionary service was delivered resulted in high cost care with no quality difference.
The current system of care drives destructive competition not competition on
value (Porter, 1999). Pro-competitive and outcome oriented care such as Value-based
purchasing (VBP), links payment directly to the quality of care provided, it moves the
focus from volume to value (Porter, 2009). Based upon similar mounting evidence
regarding the unsustainable cost trajectory and lack of association with value, CMS has
made the focus of its efforts to transform the current payment system by rewarding
providers for delivering high quality, efficient clinical care (James, 2012; Affordable Care
Act, 2010 ). Through a number of public reporting programs, demonstration projects,
pilot programs, and voluntary efforts, CMS has launched VBP initiatives in hospitals,
physician offices, nursing homes, home health services, and dialysis facilities (CMS
Hospital Pay-for-Performance Workgroup, 2007). In 2006, Congress passed the Deficit
Reduction Act of 2005 (DRA), which authorized CMS to develop a plan for VBP for
Medicare hospital services commencing FY 2009.
An early effort at incentives for publicly reporting process quality measures was
Medicare’s Reporting Hospital Quality Data for Annual Payment Update (RHQDAPU)
program. This is a pay-for-reporting (P4R) program that uses Medicare payment as an
incentive for hospitals to publicly report on the care they provide all adults, regardless of
40
payer. As originally mandated under the 2003 Medicare Modernization Act (MMA), the
RHQDAPU provision required that PPS hospitals report on a specified set of 10 clinical
performance measures in order to avoid a 0.4 percentage point reduction in their
Annual Payment Update (APU) for inpatient hospital services. This is the source of the
self-reported quality data for the Hospital Compare website, www.hospitalcompare.gov.
Payment reform has resulted in multiple reimbursement methodologies being
experimented with to identify provider preference and reduced cost. These include
bundled payments programs (innovation.cms.gov/initiatives/bundled-payments) where a
group of providers split a single payment by episode of care. Examples include joint
replacement or cardiac surgery have shown promise (VanLare, 2012). The Geisinger
coronary bypass program, known as ProvenCare, is designed as a flat fee payment for
surgery and all related care for 90 days after discharge. At Geisinger health care
system, these programs demonstrated a 10% reduction in readmissions, shorter
average length of stay, and reduced hospital charges. Perhaps most importantly, the
program achieved a 44% drop in readmissions over a course of 18 months (Bertko,
2010).
When P4P objectives are aligned with national best practice evidence as in the
CMS/Premier Quality Incentive Demonstration (HQID), process quality measures
improved significantly over a matched control group of non-participating hospitals
(Lindenauer, 2007). The study enrolled 266 participants in the HQID who were
matched with 406 Hospital Quality Alliance (HQA) controls. Hospitals needed to have a
minimum of 30 cases per condition (heart failure, heart attack, pneumonia, cardiac and
41
orthopedic surgery measures) annually to be eligible for the demonstration. For each of
the clinical conditions, hospitals performing in the top decile on a composite measure of
quality for a given year received a 2% bonus payment in addition to the usual Medicare
reimbursement rate. Hospitals in the second decile received a 1% bonus. Bonuses
averaged $71,960 per year and ranged from $914 to $847,227. A participation
requirement was that all hospitals accept a risk of financial penalty. These penalties
ranged from 1 to 2% of Medicare payments for the conditions under evaluation. They
applied if by the end of the third year of the program they ranked in the lowest two
deciles of hospitals. This is one of the earlier programs in which providers
demonstrated a willingness to accept risk based agreements (CMS, VBP, 2010) this is
an important but largely forgotten fact.
Gain sharing, once controversial but now in popular use, is another approach to
reward incentivized behavior. Gainsharing arrangements, particularly those used in
hospital and integrated delivery systems, provide bonus payments to physicians and
other providers, to reward cost savings resulting from their efforts to reorganize delivery
of clinically appropriate care at a lower cost (Eldridge, 2011). The benefit for Medicare
is that CMS shares in 50% of the reduced payments. Accountable Care Organization
(ACO) type programs, where total cost of care for a population is assigned to the
participating practitioners, has now evolved more towards a value based paradigm. In
the new ACO/gain sharing scenarios providers now have down side risk if they
overspend the funds allocated to their population which often is calculated as the prior 3
42
year average per capita spend (https://www.cms.gov/Medicare/Medicare-Fee-for-
Service-Payment/ACO/index.html).
Regardless of the foregoing, Porter (2006) is critical of payment reform
methodologies such as P4P which do not seek to keep the focus on value. He criticizes
their emphasis on compliance with evidence based guidelines and algorithms because
they lack focus on outcomes of care such as readmission, mortality, cost per discharge,
etc. More important is that they do not discourage excess utilization. Donabedian
described clinicians as maximalist when their approaches to incremental care result in
little change in outcome, Porter expressly discourages reward for P4P in these
programs. A financing incentive linked to process measure improvement, P4P, rewards
providers for achieving pre-established performance objectives in defined medical
conditions and procedures. However there is some evidence that at least process
quality improved in at least one large national study (Lindenauer, 2007).
PROCESS VERSUS OUTCOME - ASSESSING QUALITY IN THE MEANINGFUL USE
& VALUE BASED ERA
The correct method to select in order to objectively assess quality and cost
outcomes, the two basic components of the value equation, remain very much in
debate. The current practice for evaluating EMR impact on healthcare is the analysis of
its impact on cost and quality. The most widely applied method to evaluate quality
performance is the use of process measures (Donabedian, 1988; Mant, 2001; Mainz,
43
2003; Lilford, 2007). However, as Pronovost (2004) notes, the IOM defines quality as
“the degree to which health services for individuals and populations increase the
likelihood of desired health outcomes (emphasis added) and are consistent with current
professional knowledge.” As CMS enters its third year of the Value Based Payment
Program for Medicare (CMS, 2010) the focus is clearly on cost and outcome measures.
The emphasis on process measures has diminished from 60% to 10% of the
reimbursement methodology. Numerous proponents (VanLare, 2012; Ranawat, 2009;
Porter, 2009) of the outcome measures emphasize the need to shift from mere
adherence to standards, to risk adjusted outcomes of care where value not volume is
the driver.
However, from the clinical standpoint Liford (2007) and Mant (2001) support the
contrary premise that process measures are direct measures of the quality of health
care, provided that a link has been demonstrated between a given process and
outcome. In addition, they advocate for process measures because they are well
defined, easily measured, sensitive, specific and easy to interpret. Their construct
validity is derived from professional societies such as American Heart Association,
American College of Cardiology, etc. and therefore well vetted and form the community
standard of practice. The controversy of process versus outcome measures is a
consistent theme of quality literature (Pronovost, 2004; Mant, 2001; Liford, 2007). The
acuity of patient’s conditions, their compliance with care plans, their financial and
sociologic backgrounds, referred to as social determinants of health, all influence
important outcomes such as readmission, mortality and control of chronic disease. In
44
these circumstances it would be as Mant (2001) notes a “misnomer” to refer to outcome
measures as performance indicators since this would indicate a “barometer” of
population health. In an insightful and pertinent reference to the usefulness of outcome
measures, Mant notes outcome data should be used to “inform upon wider aspects of
health policy”, which Value Based Purchasing is a central theme. (p. 479)
The current reliance on process measures is not based solely upon concerns
with the potentially confounding factors referenced above. Experts in quality have
postulated that the triad of structure-process-outcome forms the foundation of quality
improvement processes. Donabedian (1966, 1988), a practicing physician,
emphasized the importance of both process and outcome measures. He stated “we
cannot claim either for measurement of process or measurement of outcome an
inherently superior validity, since the validity of either flows to an equal degree from the
validity of the science that postulates a linkage between the two. But process and
outcome do have, on the whole, some different properties that make them more or less
suitable objects of measurement for given purposes” (p. 1746).
Mainz (2003), also supports a strategy to employ both process and outcome
measures depending on the purpose. Process measures are especially useful when
quality improvement initiatives are being initiated as they can be applied to small
samples and are sensitive to small differences making them desirable for departmental
and local analysis. The foundation of the application of these measures is that they be
valid, requiring rigorous testing to produce the desired outcome, which he refers to as
“outcome validated”, and therefore represent direct measures of quality.
45
Outcome measures are more suited to assess the effectiveness of process and
as end points of care (Lilford, 2007; Mant, 2001). Examples of hospital specific
outcome measures include mortality, infection and readmission rates. However,
outcome measures have the added feature of being able to encompass broad public
health measures such as cancer, influenza and cardiac disease prevalence. As Mainz
(2003) notes, it can be recommended that the broader the perspective required, the
greater the relevance of outcome indicators.” (p. 527) It is therefore not surprising that
CMS, as the largest payer of healthcare in the United States, has moved towards the
implementation of outcome measures as a yardstick for quality performance and
payment.
Table III: Examples of Indicators Related to Structure, Process & Outcome
© 2016 Joseph G. Conte
46
Illustrated in Table III are indicators for conditions such as stroke, asthma, heart
attack care, women’s health and ICU care. What can be discerned from each of the
indicators is the continuity between the structural component, the process to be
employed and the desired outcome to be achieved. As Mainz (2003) astutely
observed, process measures are especially usefully for evaluating compliance with
standards of care and feedback on departmental performance. The outcome measures
on the other hand, are more a check on the effectiveness of the implementation of the
standards and broader public health or policy objectives that are desired to be achieved.
CMS, the largest payer of health care in the three trillion dollar U.S. health system, has
evolved from a passive payer of services to a more demanding consumer seeking to
achieve a balance between cost and quality; hence the advent of Value Based
Purchasing Program (Federal Register, 2011).
SUMMARY OF THE LITERATURE TO DATE
There are three distinct but related areas of literature being utilized to inform the
parameters of study on EHR impact on quality cost and outcome. They are literature
that assess the impact EHR adoption has on healthcare performance, literature that
utilizes differing measurement philosophies to measure performance (process vs
outcome) and the literature on the payment reform programs.
While the need to improve quality and safety outcomes is of paramount
importance (IOM, 1999; IOM 2001) the $3 trillion cost associated with the U.S.
healthcare systems is an unsustainable financial burden for the government (Porter,
2006; Darling, 2010)) and individuals (Polsky, 2009). The need to clearly identify what,
47
if any, cost savings can be associated with the adoption of EMR under MU guidelines,
has therefore received great attention. The review of the payment reform literature
details the shift away from fee for service or volume based care to payment for value
(Porter, 2006; Ranawat, 2009; Eldridge. 2011; James, 2012; VanLare, 2012). One
focus of this study will be to identify what if any connection exists between EHR
adoption and improved cost. As previously noted numerous studies (Himmelstein,
2009; DesRoches, 2010; Appari, 2009; Agha, 2011) have demonstrated that EMR
adoption while having generally positive impact on process measures of quality have
negative to neutral outcomes on cost of care.
GAPS IN THE LITERATURE
To make the connection between EHR adoption, MU standards and shift towards
value and improved outcomes a measurement redesign is required. The underpinning
of the EHR literature analysis conducted to date has been the reliance on process
measures to assess performance difference on quality, safety and cost (Chaudry, 2006;
DesRoches, 2010: Buntin, 2011; Jones, 2014). Whether compared within or between
hospitals before and after EHR adoption or between hospitals that have or have not
adopted the technology, reliance on process measures has been the standard
measurement rubric. This method while accepted professionally (Donabedian, 1988;
Mant, 2001; Mainz, 2003; Lilford, 2007) is in direct conflict with the measurement
paradigm used for the various VBP payment models. Since the parallel lever for
healthcare transformation, payment reform, is being simultaneously implemented with
MU standards for EHR adoption the linkage with outcomes that influence direct cost
48
such as cost per discharge, readmission, preventable ER use, complication rates, etc.
Further the cost variable used from study to study was inconsistent with some study
using cost estimates derived from annual cost reports (Himmelstein, 2009) or financial
and operational data from the American Hospital Directory (Ding, 2011). The fact that
much of this research was conducted prior to standard MU definitions being available,
i.e. the pre-MU era, also confounded reliability of the adoption stage of EHR technology
estimated by hospitals. None of the estimates of adoption were based on the MU
criteria and all data was based on survey responses which at best resulted in 50%
response rates. This lack of consistency between studies dictates that continued
research is necessary to more closely focus a lens on how MU standards for
interoperability, provider order entry (CPOE), and decision support can improve cost
and value as organizational experience with EMR evolves (Ding, 2011).
The gaps that this study will address impact both methodological and analytical
domains. Previous studies ignored the advent of value based purchasing on the
measurement paradigm. No prior study utilized the actual achievement of MU status as
measured by CMS as a sorting method to cohort MU and non-MU hospitals. Early
approaches relied on incomplete and self-reported survey data applying a HIMSS
electronic health record functionality algorithm. Another major methodological gap this
study will bridge is the prior approach to measure quality. When assessing the impact
EHR had on clinical performance prior studies relied solely on process measures not
outcome measures. This study also bridges the gap between the prior studies that
used non-standardized cost measures.
49
The question to be explored in this research is whether achieving meaningful use
status (MU) of EHR technology is associated with achieving a more favorable
relationship between cost, safety and improved healthcare outcomes. The difference
between this study and prior research that attempted to make this connection is that this
study will use standardized cost values calculated by the CMS which risk adjust for
acuity, regional cost variation and teaching status. Therefore, all cost values will be
uniform and consistent between hospitals. In addition, all MU criteria will be judged by
the ONC criteria for MU standards so that each response is internally valid and
consistent between hospital responses. In prior research, self-reported survey data
with response rate as low as 50% were used to rank MU performance. This study will
not be affected by survey response bias since all hospitals must respond to achieve
their MU incentives and those that do not are automatically categorized as non-MU.
50
Chapter III
RESEARCH METHODS
RESEARCH DESIGN
The study is a cross-sectional, retrospective design; it employs two cohorts,
Meaningful Use (MU) vs Non-MU hospitals. This research seeks to assess the impact
of EHR adoption on publicly reported outcomes for quality, safety and cost in the value
based purchasing era. As many as half of U.S. hospitals (DesRoches, 2013) did not
have a basic EHR system as of 2012 and far fewer had attested to MU standards. The
implementation of healthcare payment reform as a component of the Affordable Care
Act (ACA) in parallel with the HITECH Act created a naturally occurring experiment in
which to study the difference between hospitals who have adopted EHR technology and
attested to MU standards versus those who have not but have been equally impacted
by the Value Based Purchasing program without attesting to MU with EHR adoption, the
independent variable. Since there was no human subjects or individual level personal
health information the Seton Hall Institutional Review Board (IRB) concluded that the
study did not fall under the requirement for IRB review (See Appendix A).
Sample
Two cohorts were created from the Center for Medicare and Medicaid Services
(CMS) Hospital database. One cohort will represented hospitals that had not attested to
the MU adoption of certified EHR technology as of 2013. The second cohort were
hospitals that had attested to MU. For this study, data on EHR adoption followed strict
51
inclusion criteria for meeting MU adoption standards promulgated by the Office of the
National Coordinator (ONC). The actual records were drawn from the CMS payment
documentation file that records which hospitals received meaningful use payments and
in what years it was paid. Therefore, there was no need for proxy mapping (Appari,
2012; Furukawa, 2010) to interpret hospital survey responses previously required in
other studies to establish whether existing EHR met MU core standards as proscribed
by the ONC.
This study utilized MU payment as the inclusion criteria for EHR adoption with
MU standards. To receive meaningful use payments, hospitals had to meet the
predetermined Office of the National Coordinator (ONC) performance criteria and then
“attest” to the technology adoption. Through December 2015, Federal payments of
$21,095,328,473 have been paid to all eligible providers with nearly $13 billion going to
acute care hospitals (https://www.cms.gov/Regulations-and-
Guidance/Legislation/EHRIncentivePrograms/Downloads/December2015_MedicareEH
RIncentivePayments.pdf).
The MU data file is the most current and accurate database of hospitals attesting
to and being verified as meeting Meaningful Use standards as well as payments being
issued based on performance validation. This current study includes a sample of 4,221
hospitals or 94% of hospitals nationwide. In comparison, the American Hospital
Association annual survey of 4,474 acute care hospitals had an IT supplement
response rate of 2,796, or 62.4%. This response rate while relatively high, results in a
52
loss of data for 1425 hospitals. As discussed in Chapter 5 this may have a material
impact on accuracy and validity of the information.
DESCRIPTION OF STUDY VARIABLES
Cost per Discharge
In order to assess the difference in cost per discharge between MU and non-MU
hospitals, this study utilized standardized cost per discharge value from the CMS
database publicly reported through Hospital Compare (www.hospitalcompate.gov).
Importantly, this data file contains cost per discharge adjusted for unique characteristics
of hospitals, historically identified as confounding variables preventing meaningful cost
comparisons. The formula adjusts for differences among hospitals in geographical
location and in certain hospital-specific attributes. The latter include higher costs of
carrying on an approved teaching program, higher costs of care associated with a payer
mix that includes a higher percentage of low-income Medicare and Medicaid patient
populations, and special pass-through payments for unusual capital and other costs.
This study utilized the specific hospital data to conduct a MU vs Non-MU cohort level
comparison to assess cost impact in the value based purchasing context, no other study
utilized this approach.
Quality Outcomes
Under the Value Based Purchasing Program (Federal Register, 2011), CMS has
tied performance on clinical outcomes to reimbursement. The methodology employed
in measuring clinical quality performance in this study is the analysis of risk adjusted
mortality and readmission rates for three clinical conditions developed by CMS for
53
national reporting purposes from the outset of the program. The clinical conditions
measured include: heart failure, community acquired pneumonia (pneumonia) and
myocardial infarction (heart attack). In order to introduce this evolving approach to
quality measurement, an underpinning is required to attempt to control the confounding
factors and risk adjustment. The CMS database utilized to report facility outcomes has
adopted a risk adjustment methodology that creates a level playing field for
organizations to be compared to one another on this important outcome metric (Pitches,
2007; Roberson, 2015).
Safety Indicator
This study utilized Patient Safety Indicator 90 (PSI 90) an Agency for Healthcare
Research and Quality composite value utilized by Hospital Compare website as a safety
measure proxy measurement (www.qualityindicators.ahrq.gov/Modules/psi_resources).
Importantly for comparison purposes, the measures of serious complications reported
on Hospital Compare are risk adjusted to account for differences in hospital patients’
characteristics. The rate for each PSI is calculated by dividing the actual number of
outcomes at each hospital by the number of eligible discharges for that measure at
each hospital, multiplied by 1,000. The composite value reported on Hospital Compare
(https://www.medicare.gov/hospitalcompare/data/serious-complications.html) is the
weighted averages of the component indicators.
Hospital Demographics
Endogenous variables that are associated with hospital performance (CMS,
2014; Lin, 2014; Appari, 2012) were identified for each hospital included in the analysis.
54
These variables include hospital teaching status, hospital ownership category, and
acuity of care, as measured by the case mix index (CMI), hospital region, and activity
level as measured by annual discharges. Categorized by their MU status, each hospital
individually and the respective hospital cohort (MU vs non-MU) performance were
measured against its own performance for the baseline period 2009 versus 2013.
DATA COLLECTION
The analytic sample was comprised of 4,221 non-federal acute care hospital U.S.
hospitals using data reported from 2011 through 2013. Data was drawn from three
publicly reported national databases with respect to hospital’s technology status, costs
and performance on publicly reported clinical outcomes, functional characteristics and
demographics.
These databases are the only source utilized by CMS for reimbursement and
public reporting purposes when determining which organization had achieved EHR
implementation that meets MU standards, risk adjusted quality outcomes and
identification of standardized costs controlling for multiple variables. The majority of
data utilized on the previous assessment of EHR adoption and its impact on quality,
safety and cost by other large national studies (Appari, 2010; Ding, 2011; Agha, 2011,
Himmelstein, 2009; Jones, 2014), relied either on self-reported survey data, non-risk
adjusted clinical performance and cost report data that was not standardized for multiple
hospital specific or regional variables.
55
DATA ANALYSIS
Measurement Methodology
The Hospitals were assigned to control and treatment cohorts based upon their
EMR adoption status. The respective hospitals unique Common Identification Number
(CIN) number was used as a linking code to compile the information accurately from the
3 publicly reported databases utilized for the study. Each hospital’s publicly reported
performance data referred to above, for the respective pre and post MU attestation
periods, was be obtained.
Due to the large number of hospitals, over 4000, wide geographic dispersion and
other disparate attributes, the subject hospitals vary widely in numerous ways. To
identify and report on these variables each hospitals’ demographics profile information,
most frequently associated with likelihood to adopt EHR technology was identified from
the respective publicly reported data bases. These variables include: teaching status
which identifies if the hospital trains residents, acuity which is measured as a function of
case mix index, discharge volume measured as Medicare discharges and geographic
regions. The hospital region was coded numerical as 1-4, to comply with the CMS
methodology for identifying hospitals.
56
Figure 5. The Four Medicare Regions
Source: https://www.cms.gov/About-CMS/Agency-
Information/RegionalOffices/RegionalMap.html
The clinical outcome, safety and cost values all were risk adjusted to account for
variations age, sex, severity of patient condition (CMI), indirect medical education cost
associated with teaching status, operating expense associated with geographic location
and payments for treating uninsured known as disproportionate share or DISH
payments.
As stated above, all of the outcome measures selected were risk adjusted by the
respective agency that reported the data, thereby normalizing the values across
hospitals. For the clinical outcomes the 3M risk adjustment methodology was utilized by
57
CMS. For the standardized cost per discharge CMS developed an internal cost
adjustment methodology in conjunction with statistical experts from Acumen LLC
(http://www.qualitynet.org/dcs) and the Agency for Health care Research and Quality
(AHRQ) devised the safety composite score approach
(http://qualityindicators.ahrq.gov/Downloads/Modules/PSI/PSI_Composite_Developmen
t.pdf.)
DATA ANALYSIS METHODS
This study used the publicly reported data available through the Office of the
National Coordinator for MU status as well as the CMS clinical, cost and safety data set
(https://data.medicare.gov/data/hospital-compare). It separated the hospitals into MU
and non-MU status and then combined each of the 4221 hospital’s risk adjusted
outcome data and categorical variables into the master data set resulted in over
287,000 data elements for analysis. This robust data base combined with the risk
adjustment scheme for the outcome indicators supported a unique and detailed
statistical analysis of the difference in performance between MU and non-MU hospitals.
The statistical analysis was conducted on the outcomes of the two independent
cohorts, MU and non-MU to establish whether there a difference between hospitals that
implemented EMR and those that did not on important outcome variables. The data
was statistically analyzed utilizing SPSS version 22. The analysis included: Levene's
Test of Equality of Error Variances; Tests of Between-Subjects Effects; T-Test. All of
the data was analyzed at a minimum alpha of at least 0.05.
58
Based upon these parameters a T-test is the appropriate statistical test. It satisfies
the following criteria:
Assumption #1: The dependent variable is be measured on a continuous scale
Assumption #2: The independent variable consists of two categorical groups.
The data analysis methodology includes both descriptive statistics and inferential
statistical analysis. Descriptive statistics in the form of frequencies, means, medians
and standard deviations were constructed and utilized to examine specific
characteristics of the hospital research population. There is one independent variable
- MU with two categories and eight dependent variables, analyzed separately. There are
8 dependent variables: 3 readmission rates (heart attack, heart failure and pneumonia)
and 3 mortality rates (heart attack, heart failure and pneumonia) a safety measure and
the cost per hospital discharge.
There are five descriptive variables for the sample (Teaching, region, ownership, acuity
and number of discharges). The five categorical variables are not integrated into the
research design.
This study was the first to gather “big data” utilizing publicly reported information
which was not reliant on voluntary survey responses, included a standardized cost per
discharge metric, without being reliant on a proxy measures gathered via voluntary
survey responses to identify MU status. Therefore, the results presented in the next
chapter utilize a new lens with which to determine how electronic health records with
MU capabilities impact cost, quality and safety.
59
Chapter IV
RESULTS
INTRODUCTION
The purpose of this study was to evaluate whether there is a difference in
hospital performance between organizations that have adopted meaningful use of
electronic health records and those that have not. This chapter focuses on the
statistical analysis of the data assembled on the outcomes of 4221 hospitals.
CHARACTERISTICS OF THE SAMPLE
After the application of exclusion criteria, data was assembled on 4221 hospitals.
There were 560 hospitals eliminated from the study because they had less than 50
discharges per year, or less than at least 30 discharges per category of clinical
performance. Outcome measures were identified from publicly reported data sources,
the performance year selected for study was 2013.
The profile of a hospital is comprised of demographic and operational
characteristics. These characteristics or endogenous variables (CMS, 2014; Lin, 2014;
Appari, 2012) were identified for each hospital included in the analysis. The variables
include hospital teaching status, hospital ownership category, acuity of care as
measured by case mix index (CMI), hospital region which was coded consistent with
CMS regions and labeled 1-4 depending on state geography and activity level as
measured by annual discharges. In Table IV. the teaching status, region and
60
ownership are illustrated for each of these characteristics in the respective cohorts, MU
and non-MU.
Table IV: Frequencies and Percentage of Total by Categorical Variables
In the MU cohort, 2315 or 55% of hospitals had attested to MU by 2013.
Seventy-one % or 1637 were non-teaching facilities, with the majority (40%) located in
the South, followed by the Midwest, 29%, West and Northeast at 16% each. The
predominant ownership model was 61% voluntary, not for profit status, followed by
governmental 23%, proprietary 18% and physician owned 1.5%.
For the non-MU cohort, 1906 or 45%, had yet to attest to MU. Seventy eight %
or 1490 were non-teaching facilities, with the majority (35%) located in the South,
61
followed by the Midwest, 30%, West 21% and Northeast at 11%. The predominant
ownership model was 60% voluntary, not for profit status, followed by governmental
21%, proprietary 15% and physician owned 0.5%.
DEPENDENT VARIABLE RESULTS
The differences in outcome performance between meaningful use (MU) and non-
meaningful use (Non-MU) hospitals were analyzed through T-Test. The level of
significance utilized was P =.05. The dependent variables results describe mortality and
readmission rates for heart failure, heart attack and pneumonia, cost per discharge and
the AHRQ aggregate safety score.
In Table V. the number of hospitals who reported by condition and the mean
performance of the respective dependent variables is illustrated for all hospitals, MU
and non-MU cohorts. The “N” of each subset is a function of how many hospitals met
reporting criteria per variable. Minimum reporting thresholds were 30 discharges
annually per condition. What should be highlighted is that the difference in mean
performance in mortality is consistently in favor of the MU hospitals, as is the difference
in cost per discharge. The readmission rate is lower for all three conditions in favor of
the non-MU hospitals. It must be noted that the readmission data is calculated as all-
cause readmission. Therefore, readmission is not directly tied to the condition for which
the patient was initially discharged from the hospital. The PSI 90 or AHRQ safety score
is identical between cohorts. Further analysis of each variable and the statistical
significance of the differences between cohorts will be described later in this chapter.
62
Table V: Mean Performance Data by Care, Cost and Safety Variables
In Table VI. the summary of the T-test results are depicted. The mortality rates
are listed consecutively for the 3 clinical conditions of interest, heart attack, heart failure
and pneumonia. A statistically significant difference in favor of the MU hospitals was
identified for each condition. For the readmission measure a statistically significant
difference was found in favor of the non-MU hospitals. With respect to cost, MU
hospitals had a difference of $327 less per discharge using the CMS standardized
discharge metric. There was no difference between the hospital cohorts for the AHRQ
safety score. The following sections will describe in detail the clinical, cost and safety
results for each condition illustrated in Table VI.
63
Table VI: T-test Results on Quality Variables by Meaningful Use Status
RESULTS BY DEPENDENT VARIABLE
The following analysis is ordered according to the research questions initialy presented
in Chaper I.
Heart Attack Mortality and Readmission
The research question was: Is there a difference in clinical outcomes between
hospitals that have achieved Meaningful Use (MU) for their EHR systems verse
hospitals that have not achieved MU as measured by mortality and readmission rate for
Heart attack?
These results are reported from the CMS data base for Medicare discharges with
a primary mortality cause (cause of death) of heart attack. Medicare heart attack
mortality rates were aggregated from 2510 hospitals that reported data. As illustrated in
64
Table VII and Figures 6 and 7, the overall national rate was 15.14%. The Non-MU
hospital rate was 15.21% versus the MU hospital rate of 15.11%. This study found that
there was a statisistically significant difference was in favor of the MU hospital (P<.041).
A functional illustration of the implication of this finding would be its impact on
overall deaths per 500 thousand admissions for the specific condition. Nationally in
2013 there were approximately 3,000,000,000 Medicare discharges for heart attack,
heart failure and pneumonia. With nearly 500,000 of these discharges for heart attack.
There was an estimated reduction of 500 deaths in this condition associated with MU.
Table VII: T-test Results on Heart Attack Mortality by Meaningful Use Status
65
Figure 6. Histogram of Heart Attack Mortality Rate 2013 – Hospitals (MU=NO)
Figure 7. Histogram of Heart Attack Mortality Rate 2013 – Hospitals (MU=YES)
66
The following readmission results are reported from the CMS data base for
Medicare discharges. The patient has to have an initial or index admission of heart of
heart attack. The case was categorized as readmission if the patient was readmitted
within 30 days from the index admission with any diagnosis. This measurement is know
as all-cause readmission rate and is how CMS calculates the metric.
There were 2238 hospitals that reported on heart attack readmission rates, 1426
were from the MU cohort amd 812 from the non-MU cohort. As illustrated in Table VIII
and Figures 8 and 9, the overall mean readmission rate was 18.31% with non-MU
hospitals reporting a lower overall rate of 18.23 % versus the MU hospital rate of
18.35%. The difference was statisistically significant in favor of the non-MU hospital
(P<.011). In this condition with approximately 110,000 heart attack readmissions
nationally at least 300 readmission were avoided.
Table VIII: T-test Results on Heart Attack Readmission by Meaningful Use Status
.
67
Figure 8. Histogram of Heart Attack Readmit Rate 2013 – Hospitals (MU=NO)
Figure 9. Histogram of Heart Failure Readmit Rate 2013 – Hospitals (MU=YES)
68
Heart Failure Mortality and Readmission
The research question was: Is there a difference in clinical outcomes between
hospitals that have achieved Meaningful Use (MU) for their EHR systems verse
hospitals that have not achieved MU as measured by mortality and readmission rate for
heart failure?
These results are reported from the CMS data base for Medicare discharges with
a primary mortality cause (cause of death) of heart failure during the admission. Heart
failure mortality rates from 3625 hospitals were reported. As illustrated in Table IX and
Figures 10 and 11, the overall national mortality rate was 11.81%. The Non-MU
hospital rate was 11.90% versus the MU hospital rate of 11.75%. The difference was
statistically significant in favor of the MU hospitals (P<.003). With approximatelt
1,200,000 heart failure discharges annually the reduced mortality associated with the
benefit of MU adoption is approximately 2000 lives.
Table IX: T-test Results on Heart Failure Mortality by Meaningful Use Status
69
Figure 10. Histogram of Heart Failure Mortality Rate 2013 – Hospitals (MU=NO)
Figure 11. Histogram of Heart Failure Mortality Rate 2013 – Hospitals (MU=YES)
70
These results are reported from the CMS data base for Medicare discharges with
an index admission of heart of heart failure. The patient was categorized as
readmission if they were reamitted within 30 days from the index admission with any
diagnosis. This measurement is know as all cause readmission rate and is how CMS
calculates the metric. In the heart failure readmission analysis there were 3693
hospitals reporting data, 2156 were MU hospitals and 1538 were non-MU hospitals. As
illustrated in Table X and Figures 12 and 13, the overall national mean readmission rate
was 23.06 % with non-MU hospitals reporting a lower overall rate of 23.00 % versus the
MU hospital rate of 23.10%. The difference was statisistically significant in favor of the
non-MU hospital (P<.048). The lower readmission rate was associated with
approximately 330 less readmissions.
Table X: T-test Results for Heart Failure Readmission by Meaningful Use Status
71
Figure 12. Histogram of Heart Failure Readmit Rate 2013 – Hospitals (MU=NO)
Figure 13. Histogram of Heart Failure Mortality Rate 2013 – Hospitals (MU=YES)
72
The research question was: Is there a difference in clinical outcomes between
hospitals that have achieved Meaningful Use (MU) for their EHR systems verse
hospitals that have not achieved MU as measured by mortality and readmission rate for
Pneumonia?
These results are reported from the CMS data base for Medicare discharges with
a primary mortality cause (cause of death) of pneumonia. Pneumonia mortality rates
were reported 3888 hospitals. As illustrated in Table XI and Figures 14 and 15, the
overall national mortality rate was 12.02%. The Non-MU hospital rate was 12.14%
versus the MU hospital rate of 12.04%. The difference was statistically significant in
favor of the MU hospital (P<.000). With over 1,200,000 pneumonia discharges
annually the reduced mortality associated with MU adoption is approximately 2400 lives.
Table XI: T-test Results on Pneumonia Mortality by Meaningful Use Status
73
Figure 14. Histogram of Pneumonia Mortality Rate 2013 – Hospitals (MU=NO)
Figure 15. Histogram of Pneumonia Mortality Rate 2013 – Hospitals (MU=YES)
74
These results are reported from the CMS data base for Medicare discharges with
a primary admission cause of pneumonia. The patient was categorized as readmission
if they were reamitted within 30 days from the index pneumonia admission with any
diagnosis. This measurement is know as all cause readmission rate and is how CMS
calculates the metric.
In the pneumonia readmission analysis there were 3900 hospitals reporting data,
2221 were MU hospitals and 1679 were non-MU hospitals. As illustrated in Table XII
and Figures 16 and 17, the overall mean readmission rate was 17.61 % with non-MU
hospitals reporting a lower overall rate of 17.54 % versus the MU hospital rate of
17.66%. The difference was statisistically significant in favor of the non-MU hospital
(P<.004).
Table XII: T-test Results on Pneumonia Readmission by Meaningful Use Status
75
Figure 16. Histogram of Pneumonia Readmit Rate 2013 – Hospitals (MU=NO)
Figure 17. Histogram of Pneumonia Readmit Rate 2013 – Hospitals (MU=YES)
76
Patient Safety Composite Score
The research questions was: Is there a difference in clinical outcomes between
hospitals that have achieved Meaningful Use (MU) for their EHR systems verse
hospitals that have not achieved MU as measured by Agency for Healthcare Research
and Quality Patient Safety Indicator Measurement ?
For AHRQ composite safety score, 3163 hospitals reported data. As illustrated
in Table XIII and Figures 18 and 19, there were 1960 MU hospitals reorpting an overall
score of 0.60 and 1203 non-MU hospitals reporting and identical score of 0.60. there
was no statstically significant difference between the MU and non-MU hospital
outcomes.
Table XIII: T-test Results on Composite Safety Score by Meaningful Use Status
77
Figure 18. Histogram of Safety – Hospitals (MU=NO)
Figure 19. Histogram of Safety – Hospitals (MU=YES)
78
COST PER DISCHARGE
The research questions was: Is there a difference in cost per discharge between
hospitals that have achieved Meaningful Use (MU) for their EHR systems verse
hospitals that have not achieved as measured by the CMS standardized cost per
discharge metric?
For the standardized cost per discharge there were 31634 hospitals reporting
data. As illustrated in Table XIV and Figures 20 and 21, the national mean Medicare
discharge cost was $7975. There were 1955 MU hospitals reporting with a mean
discharge cost of $7852. There were 1179 non-MU hospitals reporting with a cost per
discharge of $8179. The difference in mean cost per discharge was $327 in favor of
MU hospitals which was statistically significant (P<.000). With over 20,000,000
Medicare discharges annualy an estimated cost reduction associated with MU is over
$6 billion. This very significant finding, the implicatiions of the cost of EHR adoption and
potential future trajectory of savings associated with MU will be discussed in detail in
Chapter 5.
Table XIV: T-test Results on Standardized Cost Per Discharge Metric by Meaningful Use Status
79
Figure 20. Histogram of Cost per Discharge – Hospitals (MU=NO)
80
Figure 21. Histogram of Cost per Discharge – Hospitals (MU=YES)
SUMMARY
There are favorable mean scores for the Meaningul Use hospitals for heart
failure, heart attack and pneumonia mortality. In addition, the average standarized cost
per discharge is lower for MU hospitals by $327. There is no difference in mean safety
score (PSI 90) between hospital cohorts. The readmission results reveal that non-MU
hospital had lower all-cause readmission rates in all three clinical domains. While the
differences are not large in comparing the raw rates when assessing the difference
based upon the number of discharges impacted the number of lives saved and cost
81
reduced is substantial. From the cost perspective wiith the cost per discharge
difference of $327 and the 20,204,517 discharges included in the research, the dollars
saved amount to over $6.6 billion. In lives saved the data is likewise impactful. When
applying the improved mortality rate to the over 20,000,000 Medicare discharges from
the nation’s hospitals a reduction in mortality of over 20,000 lives is associated with MU
adoption. The finding regarding readmission rates in favor of non-MU hospitals was
unexpected. The phase1 MU guidelines are substantially focused on inter-facility
integration and interoperability. As the phase 2 MU guidelines take hold with their focus
on care plan integration, E-prescribing and related data sharing with external, non-
hospital providers, a positive impact on readmission rates is expected.
Table XV illustrates the results for the Levene’s test. The equality of variance
test results were accepted for heart attack readmit and heart attack mortality rates,
heart failure and pneumonia mortality rates. The equality of variances results were
rejected for heart failure and pneumonia readmission rates, safety and cost measures.
In the cases where the equality of variance was rejected the alternate degrees of
freedom and t-test scores were utilized to properly calculate statistical significance of
the measurements.
82
Table XV: Levene’s Test for Equality of Variance of Quality Variables (2 sample t-test)
Levene's Test for Equality of
Variances
F Sig.
Heart Attack Readmit Rate .939 .333*
Heart Attack Mortality Rate .310 .578*
Heart Failure Readmit Rate 17.484 .000^
Heart Failure Mortality Rate 1.771 .183*
Pneumonia Readmit Rate 9.830 .002^
Pneumonia Mortality Rate .181 .670*
Safety 9.093 .003^
Cost per Discharge 45.419 .000^
*equality of variance assumed ^ equality of variance rejected
+Equal variances not assumed for the t-test
++Equal variances assumed for the t-test
*p<.05 **p<.01
83
Chapter V
DISCUSSION
BACKGROUND
From a health policy perspective, the $50 billion CMS committed to the HITECH
Act is a substantial investment in the implicit belief that EHR adoption will transform the
U.S. healthcare system. While the program’s stated purpose was to support and
stimulate the adoption of EHRs’ in healthcare, addressing the value inequity between
cost and outcomes in the $3 trillion U.S. healthcare system is an outcome of great
interest (IOM, 2001). Simultaneously, Value Based Purchasing, VBP, is realigning the
reimbursement paradigm by shifting payments from fee for service to payment for value.
In this scheme, outcome and cost are the respective numerator and denominator to
measure value, inexorably linking these two initiatives (Porter, 2006).
Early studies undertaken to assess EHR impact on clinical and financial
outcomes were primarily undertaken in the pre-MU era (Chaudry, 2006; Ding, 2011;
Himmelstein, 2009; DesRoches, 2010). Those studies used process measures to
assess impact and cost estimates were generally derived from various sources such as
cost reports, AHA survey responses, financial filings, etc. The assessments of safety
were generally focused on hospital centric studies on important indicators such as
medication error, falls and infection rates (Bates, 2010; Poon, 2010). Due to the
complexity of identifying these outcomes from administrative data these results were
difficult to validate and replicate across large numbers of organizations. The need to
84
validate the interim progress that MU has on clinical outcomes and cost requires a
revised measurement paradigm aligned with Value Based purchasing concepts of
outcome and cost (Porter, 2006) with a global focus on patient safety (AHRQ, 2006).
The purpose of this research was to evaluate whether there is a difference in
hospital performance outcomes in organizations that have implemented electronic
health records that meet Meaningful Use (MU) standards. The use of a revised
measures paradigm, one focused on publicly reported outcome measures, not process
indicators, is in alignment with payment reform under the Affordable Care Act. The
outcomes of interest, as stated, were mortality and readmission rates, cost per
discharge and aggregate safety score. With a national healthcare bill of over $3 trillion,
the American healthcare system spends nearly double the amount of every
industrialized nation on a per citizen basis. Ironically, the U.S. has the lowest life
expectancy and the highest infant mortality rate of the group (OECD, 2014). In addition,
the Institute of Medicine estimates that the third leading cause of death in America is
related to patient safety lapses (Squire, 2012). The need for a realignment of cost,
outcome and patient safety is of paramount importance.
This study sought to ascertain whether there is a relationship between
Meaningful Use of EHRs’ and quality, cost and safety outcome measures. In reviewing
the study findings, after analyzing outcome data on 4221 hospitals the conclusion is that
there is a statistical difference for mortality rates for all three conditions for meaningful
use hospitals; heart attack, heart failure and pneumonia. Further, meaningful use
hospitals demonstrated statistically significant difference in terms of standardized cost
85
per discharge. As discussed in greater detail later in this chapter, when extrapolating
the mortality difference achieved at MU hospitals there were over 20,000 lives saved.
This is a powerful finding and one further magnified when taken in the context that over
$6 billion in cost per discharge was achieved by the same cohort (MU).
These results are the first definitive endorsement of MU capability in clinical
quality and cost savings. When considering that the lead time for adopting complex
technology is estimated at between two and five years (Ding, 2011), the fact that
mortality differences and cost savings were demonstrated in the first two years after the
initial attestation period, 2013, is support of the CMS investment. Other potential
explanations for outcomes improvements aside from MU implementation will be
discussed later in this chapter.
DISCUSSION
Impact of the HITECH Act
The implementation of basic EHRs’ in the nation’s hospitals stood at just 8% in
2008 (Jha, 2009). With the passage of the HITECH Act in 2009, ushering in both
monetary incentives and penalties for EHR adoption that had to meet MU standards, a
veritable rush for implementation impacted the healthcare industry. In fact by the time
the first incentive payments were available in 2010-2011 period there was an over
threefold increase of EHR adoption to nearly 27% of hospitals (DesRoches, 2013). The
initial uptake, according to the American Hospital Association annual IT survey, was in
large, urban and teaching hospitals. The research suggested that this hospital cohort,
86
large, urban, teaching hospitals, was twice as likely to have adopted an EMR and that
approximately 44% could meet MU standards (DesRoches, 2013). The AHA survey
response rate at just over 60% did not include the many hospitals, at least 1500, and
the actual results from this study discussed below differed materially.
By analyzing the actual 2013 MU attestation data file, the current study found
that the percentage of hospitals that had actual MU certified EHRs’ had jumped from
8% to 55%, or 2315 of 4221 hospitals nationally. In contrast to the DesRoches (2013)
study, the majority of hospitals that had actual MU certified functionality, 1637 of the
2315, or 71% of the attesting cohort, were actually non-teaching facilities. The majority
(40%) located in the South, followed by the Midwest, 29%, West and Northeast at 16%
each. The predominant ownership model was 61% voluntary, not for profit status,
followed by governmental 23%, proprietary 18% and physician owned 1.5%. The
difference in the data reported by DesRoches (2013) and the actual CMS data results
just one year later may be interpreted in several ways.
One reason for the difference in actual versus reported uptake in MU certified
EHRs s that the Desroches (2013) study relied on voluntary survey data with a 61%
response rate. A large number of hospitals, over 1500, did not reply, many of whom
were likely not AHA members or have seen the value in completing the survey. To
receive MU payments it was mandatory for hospitals to attest and to be certified as MU
compliant, therefore the CMS data file used for the current study had the most current
and accurate data. The other reason, also aligned with a payment incentive, is that
hospitals clearly moved very quickly, a 7 fold uptake, to advance their basic EHR
87
capability once the MU criteria was finalized in order to capitalize on the HITECH funds.
Therefore the financial incentive seemed to have clearly increased the number of EHR
installations by those meeting the criteria. In the first year over $6 billion was awarded
to hospitals.
Clinical Quality
However, the most striking aspect of this study is that the clinical outcomes
reported reveal statistically significant difference in mortality rates in all three clinical
conditions, heart attack, heart failure and pneumonia for hospitals who achieved MU
recognition for their EHR systems. This critical finding represents both a quality and
reimbursement benefit to hospitals. The pressure to move away from fee for service
reimbursement to value, driven by the VBP model has refocused the quality discussion
to one centered on clinical outcomes and away from process measures. Previous
research (Ding, 2011; Appari, 2012; Himmelstein, 2009; Chaudry, 2006; Spencer, 2012)
did not demonstrate this level of clinical improvement either in magnitude of change or
uniformly across conditions. It is important to discuss the focus on process versus
outcome as a measurement paradigm in this prior research to understand the different
outcomes of the studies.
The community standard for measurement of clinical quality since 2003 when the
first publicly reported data by CMS (www.medicare.gov/hospitalcompare.gov) and the
Joint Commission (www.JCAHO.org) has been the utilization of aggregated process
measure data. As discussed in the literature review, individual quality experts such as
Donabedian (1988, 2003) as well as institutional authorities on quality measurement
88
such as the National Quality Forum, advocated for process measure standards.
Indeed, the theoretical support for process measures is strong (Mant, 2001; Mainz,
2003; Lilford, 2007). However, the process measure philosophy is tied to the linkage
between process and clinical standard validity and best suited for practitioner feedback
and performance improvement. However, the focus on population health inherent to
VBP requires a measurement lens of broader scope and one oriented to informing wider
aspects of health policy (Mant, 2001), that being outcomes. Outcome measures such
as mortality, readmission and infection are discrete events. By focusing on mortality
and readmission as quality endpoints, consistent with new VBP measures, this current
study was able to report results less subject to such data management concerns (Mant,
2001; Rubin, 2001).
Procedurally, the utilization of process measures requires data be abstracted
from administrative systems. This method is attractive since it is automated for large
data sets, less expensive and efficient. It does have validity issues, however, these
limitations are a function of the completeness and accuracy of the individual
documentation of each clinical intervention as transposed into the hospital record and
billing systems (Billings, 2003; Grosse, 2010; Tollefson, 2011). Each aggregate quality
measure for the clinical conditions under study has at least 8 sub processes that must
be performed and documented to achieve a “passing grade” for the clinical encounter.
The variability of the documentation and data management associated with the process
measures approach to quality measurement creates opportunity for error. The outcome
measure methodology utilized in the current study and discussed below alleviates these
89
Meaningful Use
Another issue requiring elaboration is the utilization of actual Office of National
Coordinator (ONC) criteria to create the two cohorts used for analysis. The research
reported in this study utilized the ONC database of actual MU achievement utilized for
awarding MU status and distribution of payments. The information was validated and
audited by CMS prior to awarding a MU certification or making payments. All of the
prior studies mentioned in the literature review utilized proxy measures based on either
a Health Information Management Services Society (HIMSS) or AHA Information
Technology voluntary survey instrument to establish if a hospital had implemented an
EHR that was capable of meeting MU specifications. These studies relied on self-
reported capabilities from either of the HIMSS or AHA hospital surveys. The accuracy
and response rate create a question as to the accuracy of categorizing a hospital in a
specific cohort, MU or non-MU. Further, as mentioned above in any given survey
response year a large number of hospitals, over 1500, did not reply omitting a
significant number of organizations from their analysis. The impact of the inclusion of
many non-surveyed hospitals in the current study significantly affected the true total of
organizations meeting MU standards in the first 2 years, and affected how the cohort
performed in the cost savings described below.
Cost
The fact that a statistically significant difference in cost per Medicare discharge between
MU and non-MU hospitals of $327 was found in this study provides support for the $50
billion investment of the HITECH Act. In one year, with just over 50% of the nation’s
90
hospitals participating, the current study suggests over $6 billion in savings will have
accrued from MU adoption. The focus on whether EHRs’ demonstrated a relationship
to cost in the $3 trillion U.S. healthcare system was a significant focus on this study.
The pressure for policy change to create a value driven healthcare system under the
American resource and Recovery Act (AARA) of 2009 was supported by economist
Porter (2006). CMS sought to bend the cost curve and align payment with value by
implementing VBP and shifting the original reimbursement equation weighted 90% in
favor of process to the current 2016 formula which is 90% outcome oriented
(https://www.medicare.gov/hospitalcompare/data/total-performance-scores.html).
Prior studies have utilized various methods for detecting the impact of an EHR on
hospital costs. Himmelstein (2009) used Hospital Medicare Cost Reports to assess an
organization’s overall administrative cost. Ding (2011) used American Hospital
Association survey data to create two financial indicators, operating cost per day and
operating cost per admission. DesRoches (2010) and Agha (2011) utilized Medicare
Provider Analysis and Review File and Medicare Inpatient Impact File. Other
researchers excluded cost entirely choosing to focus on quality or utilization
approaches, consistent with practice guidelines without commenting on costs (Jones,
2014; Appari, 2012). What is clear from the literature is that past studies utilized
multiple approaches and data sources, some overlapping, others unique, utilized to
assess if EHR adoption had an impact on healthcare costs. This lack of consist
measures limits the external validity of these studies on the cost domain.
91
The difference with the current study is that none of the prior research utilized a
normalized cost per discharge approach. The current study used the CMS spending by
claim file (www.data.medicare.gov/Hospital-Compare/Medicare-Hospital-Spending-by-
Claim) that calculates a normalized Medicare spending per discharge by hospital. The
multiple characteristics of a hospital’s overall cost structure, union versus non-union
staff, urban versus rural, ownership models, payer mix, teaching status create such
variability that without an adjusted cost per discharge approach there can be no
meaningful cost comparison between the cohorts, MU and non-MU hospitals. By
utilizing the CMS Medicare spending per beneficiary file cost per discharge this barrier
to cost analysis has been removed in this study. This same standardized metric
approach was utilized to assess the final and critically important safety domain.
Safety
The previous research on EHR impact on safety outcomes focused on specific
initiatives in local hospital or health systems (Poon, 2012). The seminal study To Err is
Human (1999) identifying between 44-98,000 deaths annually from errors was followed
by numerous other studies (Bates, 2001; Poon, 2010; Shekelle, 2011). As identified in
the cost per discharge issue, the study of safety and EHR impact on improving overall
results were not undertaken on broad enough levels to create an endorsement of
technology as the hoped for change agent. Some studies actually found that EHR
created its own error prone process problems and a caution flag was raised (IOM, 2012;
Sittig, 2012). Few if any broad based studies were focused on this topic because of the
92
complexity in measuring and identifying agreed upon community standard for analysis.
The AHRQ (2010) safety metric PSI 90, changed this barrier.
Yet, while this new measurement paradigm was significant in its breadth of
measurement and ability to be extracted from administrative data the results failed to
reveal a difference in outcome between MU and non-MU hospitals in this study. One
reason may be that the indicators selected for the composite score are not well aligned
with clinical interventions that EHR can specifically impact. There are 11 indicators, 9 of
which are surgery related, followed by pressure ulcers and blood stream infection. In
order to better assess the impact of EHR on patient safety a different indicator set more
effective in impacting safety issues such as medication errors, timing of antibiotic for
procedures, pneumonia care or management of sepsis (Bates, 2001; Poon, 2010)
would be more sensitive measures.
THEORETICAL IMPLICATIONS
This study drew upon two theoretical frameworks, one focused on clinical care
measurements and the other focused on healthcare economics. The Donabedian
Model focuses on the structure, process and outcome to measure quality. The second
was the Value Based Purchasing theory advanced by economist Michael Porter in his
seminal work, Redefining Healthcare (2006). The Porter model argues that the US
healthcare system’s inherently misaligned payment methodology resulted in the pursuit
of high volume, high margin services without a focus on the outcomes of health for the
population being served or the total cost of care incurred.
93
The Donabedian quality measurement theory remains a consistently applied and
valid approach to measuring quality from a process perspective. The CMS hospital
compare program, Joint Commission and National Quality Forum utilize and endorse
quality measurement at the process level. However, the measurement framework is
best applied at the practice feedback level. For example assessing performance and
giving feedback re compliance with or missed care opportunities, such as administration
of therapy within proscribed time frames such as aspirin within 60 minutes for heart
attack patient, antibiotic within 30 minutes for pneumonia patients in the ER. Mainz
states simply “process indicators assess what the provider did for the patient and how
well, it was done” (p. 525).This approach is in contrast to outcome measures with their
focus on population and endpoint measurement.
Process measures are at best useful in a Pay-for-Performance approach such as
the Premier/CMS Project (Lindenauer, 2007) that aligned payment with achieving the
highest levels of compliance with care guidelines. It was ultimately concluded that the
program improved compliance but never decreased cost, reduced safety errors, or
changes endpoints in mortality or readmission. So while the process measurement
theory remains a valid and useful tool in quality improvement efforts, for the purposes of
validating broader population measures such as mortality rates, readmission and safety
metrics the greater the relevance of outcome measures (Mainz, 2003). Most
importantly however, process measurement is methodologically unsuitable for
measuring outcomes and hence out of sync with the Value Based Purchasing
94
reimbursement paradigm implemented by CMS where 90% of current payments are
focused on outcome and not process measures.
LIMITATIONS
As with all studies, this study has several limitations. First, hospitals cannot be
randomly assigned to control and treatment groups as in a randomized control trial but
they could be assigned to cohorts based upon their EMR adoption status. Due to the
large number of hospitals, 4221, resulting in wide geographic dispersion and other
disparate attributes, the subject hospitals vary widely in numerous ways: teaching
status, urban vs rural, large vs small discharge volume, union status and related social
determinants of patients. This is a potential threat to the generalizability of the study
conclusions. To control for this threat, all of the outcome measures selected were risk
adjusted thereby normalizing the values across hospitals.
The data utilized for this study was abstracted from sources that utilized
administrative data. As discussed above there are inherent limitations to this data
source, however, outcome measures such as mortality, readmission and infection are
discrete events. By focusing on mortality and readmission as quality endpoints,
consistent with new VBP measures, this study was able to report results less subject to
such data management concerns. In addition, the 3M risk adjustment methodology
applied by CMS for the clinical outcome measures while the current standard for risk
adjustment in the industry is subject to the criticism of all such formulas (Rubin, 2001).
The standardized cost metric utilized for assessing cost per discharge was developed
by CMS. The cost is calculated from hospital specific data and then risk adjusted for a
95
number of variables, medical education, geographic cost allowances, etc. affecting a
hospital’s operating expense (https://www.cms.gov/Medicare/Quality-Initiatives-Patient-
Assessment-Instruments/hospital-value-based-purchasing/Downloads/). As such this
metric is prone to all criticisms that any risk adjusted value may be subject. The actual
cost savings calculated in this study is based upon this value and may not be directly
linked to actual reduced hospital operating expenses. It is also focused solely on
standardized hospital discharge cost not total cost of care per episode attributable to the
population.
FUTURE RESEARCH
This study was undertaken to focus a lens on EHR adoption, a major health
policy initiative under the HITECH Act (2009). The $50 billion investment was aimed at
transforming healthcare by accelerating the adoption of EHRs. Future research is
required to assess the ongoing impact EHR adoption under MU guidelines will have on
clinical, cost and safety outcomes as larger and larger numbers of hospitals meet the
requirements or face penalties. This is especially true as Meaningful Use Phase 2
places greater focus on connectivity between providers, E-prescribing, care plan
exchange, greater utilization of evidence based guidelines and patient engagement via
portal use is now being implemented.
As more care is being directed toward lower cost settings, the ambulatory care
platform will assume greater importance in overall healthcare spending. Therefore,
future research could include additional studies aimed at evaluating how and if, the
increasing adoption of EHR by non-hospital providers such as physician practices,
96
therapist offices, nursing homes, home care agencies, pharmacies and other ancillary
providers has the potential to result in increased clinical and cost benefits. Specifically,
a focused examination of how readmission within 30 days of an index hospitalization
can be impacted by EHR adoption and interoperability when the constellation of
providers mentioned above have facilitated electronic communication.
Another area requiring future research is the continued refinement of patient
safety indicators that can be measured within the EHR platform of hospitals and other
providers. This study utilized the AHRQ, PSI 90 patient safety composite score which
revealed no significant difference between the cohorts under study. Their appeared to
be low sensitivity between the indicators that comprise the score to processes that are
affected most directly by EHR functionality, i.e. medical management versus surgical
interventions.
The issue of cost continues to be a prominent one in discussing the future state
of the U.S. healthcare system. Future research that identifies the impact of specific
EHR functionality on total cost of care is required to identify, refine and expand the
functionality that maximizes the cost benefit of healthcare dollars expended.
As noted earlier, legislation affecting both the community physician practices
and hospitals under the Medicare Access and CHIP Reauthorization Act released in
April of 2016 is focused on how adoption and outcomes should be aligned with payment
reform. In the new reimbursement paradigm payment for value not volume is an
underpinning of healthcare transformation (Porter, 2006). Therefore, further research is
97
required to see how these intersecting forces, EHR systems and VBP, impact a new era
in which fee for service medicine recedes payment for value ascends.
CLOSING COMMENTS-FUTURE INDUSTRY TRENDS
The future improvement in the healthcare system will require additional
cooperation and integration between hospitals, community based providers, continuing
care organizations, home care agencies, ancillary testing providers and others. The
electronic health record information linkages between the providers, the payers, and
oversight agencies are critical to improving quality outcomes, and reducing overall cost
of care. The infrastructure of the National Health Information Network may be a
powerful tool in this pursuit.
At the center of all of this change is the patient. For change to be meaningful
and lasting, culturally competent care must be provided to patients by a competent
workforce motivated to improvement. Technology, including EHR, is a tool that can
support these efforts and the evidence assembled by this study suggests that it is a
powerful one. As discussed above, future research is required to understand the
implications of EHR in conjunction, not in isolation, of other initiatives. Improvements in
hospital outcomes of care are a national responsibility of the healthcare system from a
regulatory, professional and fiduciary perspective.
In conclusion, this study found that there is a positive difference in cost per
discharge and clinical outcomes between hospitals that have and have not adopted MU
technology in their day to day operations. As concepts of interoperability between
98
hospitals, physician practices and out-patient providers advance in the next stage of MU
implementation more gains are possible. Based on the current study over 21,000 lives
were saved and up to $6.6 billion dollars in expenditure avoided related to MU
implementation. As the remaining hospitals across the nation close the gap in adopting
EHRs’ with MU functionality further benefits may accrue if this trajectory of improvement
holds.
99
REFERENCES
Appari, A., Eric Johnson, M. and Anthony, D. L. (2012), Meaningful use of electronic health record systems and process quality of care: evidence from a panel data analysis of U.S. acute-care hospitals. Health Services Research, doi: 10.1111/j.1475-6773.2012.01448.x doi: 10.1111/j.1475-6773.2012.01448.x
Bates, D., Cohen, M., Leape, L., Overhage, M., Shabot, M., Sheridan, T. (2001). Reducing the Frequency of Errors in Medicine Using Information Technology. Journal of the American Medical Informatics Association, 8(4), 299–308.
Brenna, T., et al. Incidence of adverse events and negligence in hospitalized patients: results of the Harvard Medical Practice Study. Qual SafHealth Care 2004;13:145–152
Byrne, C., Pan, E., Vincent, A., Johnston, D., Middleton, B. (2010). The value from investments in health information technology at the U.S. department of veterans Affairs. Health Affairs, 29(5), 629-638.
Chaudry, B., Wang, J., Wu, S., Maglione, M., Mojica, W., Roth, E., Shekelle, P. (2006). Systematic review: impact of health information technology on quality, efficiency, and costs of medical care. Annals of Internal Medicine, 742-751.
Roadmap for Implementing Value Driven Healthcare in the Traditional Medicare Fee- for-Service Program. https://www.cms.gov/.../VBPRoadmap_OEA_1-16_508.pdf
Davis, K., Schoen, C.., Schoenbaum, S., Doty, M., Holmgren, A.., Kriss, J., Shea, K. (2007) Mirror, Mirror on the Wall: An International Update on the Comparative Performance of American Health Care. Common Wealth Fund. Volume 59.
Ding, D., Xiaoseng, D. (2011) Long-term effects of electronic medical records on hospital clinical, financial and operational performance. POMS
100
Conference Proceedings, Abstract number: 025-0837.
Donabedian, A. (1988) The quality of care How can it be measured? Journal American Medical Association, 1743-1748.
Federal Register. (2011) Hospital Value-Based Purchasing Program (CMS- 3239-F)
Himmelstein, D., Wright, A., Woolhandler, S. (2009). Hospital computing and the costs and quality of care: A national study. The American Journal of Medicine, 1-7.
Institute of Medicine. (1999). To Err is Human: Building a Safer Health System. Washington DC: Author.
Institute of Medicine. (2001). Crossing the Quality Chasm: A New Health System for the 21st Century. Washington DC: Author.
Jha, A., Desroches, C., Campbell, E., Donelan, K., Rao, S., Ferris, T., Blumenthal, D. (2009). Use of electronic health records in U.S. hospitals. The New England Journal of Medicine, 1628-1638.
Lilford, R., Brown, C., Nicholl, J. Use of process measures to monitor the quality of clinical practice (2007). British Medical Journal, 648-650.
Lin, Y., Lin, M., Chen, H. (2014) Beyond Adoption: Does Meaningful Use of EHR Improve Quality of Care? Social Research Science Network. http://ssm.com/abs tract=2444054
Lindenauer, P., Remus, D., Roman, S., Rothberg, M., Bratzler, D. (2007). Public reporting and pay for performance in hospital quality improvement. The New England Journal of Medicine, 486-496.
Mainz, J. (2003) Defining and classifying clinical indicators for quality
101
improvement. International Journal for Quality in Health Care, 523-530.
Mant, J. (2001). Process versus outcome indicators in the assessment of quality health care. International Journal for Quality in Health Care, 475- 480.
Poon, E., Keohane, C., Yoon, C., Ditmore, M., Bane, A., Levtzion-Korach, O., Gandhi, T. (2010). Effect of Bar-Code Technology on the Safety of Medication Administration. New England Journal of Medicine, 1698-1707.
Pronovost, P., Nolan, T., Zeger, S., Miller, M., Rubin, H. (2004) How can clinicians measure safety and quality in acute care? The Lancet, 1061- 1067.
Research-Statistics-Data-and-Systems/Statistics-Trends-and Reports/NationalHealthExpendData/Downloads/highlights.pdf; www.cms.gov.(2014)
Shekelle, P. & Jones, S. Health Information Technology: An Updated Systematic Review with a Focus on Meaningful Use Functionalities. (2014) Rand Corporation Prepared for: Office of the National Coordinator for Health Information Technology
Spencer, J., Adams, J., Schneider, E., Ringel, J., McGlynn, E. (2010) Electronic health record adoption and quality improvement in US hospitals. American Journal of Managed Care, SP-64-SP71.
Wang, S., Middleton, B., Prosser, L., Bardon, C. Spurr, C., Carchidi, P., Bates, D. (2003). A cost benefit analysis of electronic medical records in primary care. The American Journal of Medicine, 397-403.
102
APPENDIX A
Seton Hall University: Institutional Review Board (IRB) Approvals
103
104
APPENDIX B
DEFINITIONS
ARRA – American Recovery and Reinvestment Act 2009, is the parent legislation that
authorized the funds for electronic health record subsidy for the HITECH Act
AHRQ - Agency for Healthcare Research and Quality is a fully funded division of the
Department of Health and Human Services. Their stated mission is “The Agency for
Healthcare Research and Quality's (AHRQ) mission is to produce evidence to make
health care safer, higher quality, more accessible, equitable, and affordable, and to
work within the U.S. Department of Health and Human Services and with other partners
to make sure that the evidence is understood and used”
(http://www.ahrq.gov/cpi/about/mission/index.html)
CMS – Center for Medicare and Medicaid Services. CMS is a federal agency that
administers health insurance programs for 100 million Americans. CMS sponsors the
healthcare website, www.cmshospitalcompare.gov that provides a portal into healthcare
services rating hospital and provider performance.
EHR system– Electronic Health Record refers to a system of interconnected electronic
health care record platforms. These systems create a platform and repository for such
functions as physician order entry, nursing record keeping, pharmacy, radiology,
105
surgery and anesthesia charting, ER and Transport systems. The system is capable of
having a outward facing portal for patient engagement, transfer of care plans and
discharge information to providers outside the hospital such as nursing homes as well
as connecting to local and national Health Information Exchanges.
HITECH Act - Health Information Technology for Economic and Clinical Health Act was
enacted under the ARRA legislation specifically to spur adoption of EHRs. The HITECH
Act set Meaningful Use of interoperable EHRs systems as a critical national goal and
incentivized EHR adoption. Penalties for non-adoption were also a part of the program.
Interoperability – The complex US health care system is comprised of numerous
electronic health record (EHR) products. Interoperability refers to the architecture or
standards that make it possible for diverse EHR systems to work compatibly in a true
information network exchanging information between providers.
Meaningful Use - The Meaningful Use aspect of the HITECH Act is part Medicare and
Medicaid EHR Incentive Programs that sets out specific performance and compliance
criteria for providers to demonstrate that their certified EHR technology meets specific
measurement thresholds that range from recording patient information, accessing
clinical evidence, patient portal, external data transmission, syndromic surveillance
capability all as structured data.
Mortality – is a measure that calculates actual death during a hospital stay, it does not
include hospice services. For this study the data is risk adjusted, it does include
106
hospitalizations for Medicare beneficiaries 65 or older who were enrolled in Medicare for
12 months before their hospital admission.
Readmission - measures that are calculations of unplanned readmission to an acute
care hospital in the 30 days after discharge from a hospitalization. Patients may have
had an unplanned readmission for any reason. For this study the data are risk adjusted.
(https://www.medicare.gov/HospitalCompare/Data/30-day-measures.html)
Risk adjustment- To accurately compare hospital performance, the CMS readmission
and death measures adjust for patient characteristics that may make readmission or
death more likely. These characteristics include the patient’s age, past medical history,
and other diseases or conditions known as comorbidities the patient had when they
were admitted that are known to increase the patient’s chance of dying or having an
unplanned readmission.( https://www.medicare.gov/HospitalCompare/Data/30-day-
measures.html)
VBP - Hospital Value-Based Purchasing (VBP) is part of the Centers for Medicare &
Medicaid Services’ (CMS’) effort to link Medicare’s payment system to a value-based
system to improve healthcare quality, including the quality of care provided in the
inpatient hospital setting. The program attaches value-based purchasing affecting
payment for inpatient stays in over 3,500 hospitals across the country. Participating
hospitals are paid for inpatient acute care services based on the quality of care, not just
quantity of the services they provide. Congress authorized Inpatient Hospital VBP
107
under the Affordable Care Act. (https://www.cms.gov/Medicare/Quality-Initiatives-
Patient-Assessment-Instruments/hospital-value-based-purchasing/index.html)